Sound signal encoding and decoding method, device, program product and recording medium

Abstract

The downmixing unit 110 obtains a signal obtained by mixing the input audio signal of the left channel and the input audio signal of the right channel. The left channel signal subtraction unit 130 and the right channel signal subtraction unit 150 encode the difference between the input audio signal, the multiplication value of the downmixed signal and the subtraction gain for each of the left channel and the right channel. In such a structure, the left channel subtraction gain estimation unit 120 and the right channel subtraction gain estimation unit 140 determine the subtraction gain so that the quantization error generated by the two encoding / decoding becomes small.

Description

Technical Field

[0001] The present invention relates to a technology for performing embedded encoding / decoding on a two-channel sound signal. Background Art

[0002] As a technique for performing embedded encoding / decoding of a two-channel audio signal and a monaural audio signal, there is the technique of non-patent document 1. The following technique is disclosed in non-patent document 1 (refer to Figure 8 etc.): a monaural signal is obtained by adding the sound signal of the input left channel and the sound signal of the input right channel, the monaural signal is encoded (mono encoding) to obtain a monaural local decoded signal, and for each of the left channel and the right channel, the difference between the input sound signal and the monaural local decoded signal is encoded. In the technology of non-patent document 1, in the differential encoding, not only the difference between the sound signal of each channel and the monaural signal is encoded, but also the quantization error of the monaural encoding is encoded, thereby reducing the quantization error of the monaural signal contained in the decoded sound signal of each channel on the decoding side, and suppressing the sound quality degradation of the decoded sound signal of each channel.

[0003] On the other hand, as a monaural encoding method for obtaining a high-quality monaural decoded signal, there is the technology of non-patent document 2. If a high-quality monaural encoding method such as the 3GPP EVS standard of non-patent document 2 is used as the monaural encoding of non-patent document 1, embedded encoding / decoding of a two-channel audio signal and a monaural audio signal with higher sound quality can be achieved.

[0004] Prior art literature

[0005] Non-patent literature

[0006] Non-patent document 1: Bernhard Grill, Bodo Teichmann, "Scalable Joint Stereo Coding" AES 1998

[0007] Non-patent document 2: 3GPP EVS standard (3GPP TS26.445) Summary of the invention

[0008] Problems to be solved by the invention

[0009] In the monaural coding method of non-patent document 2, an algorithmic delay exceeding the frame length is required to obtain a monaural local decoded signal. If a monaural coding method such as non-patent document 2 is used as the monaural coding of non-patent document 1, the algorithmic delay for obtaining a monaural local decoded signal becomes a problem in a use method requiring low delay. In addition, since decoding processing is required in the coding device to obtain a monaural local decoded signal, the amount of computational processing for obtaining a monaural local decoded signal becomes a problem in a use method requiring low computational complexity.

[0010] Therefore, an object of the present invention is to provide embedded coding / decoding for a two-channel audio signal, which suppresses the sound quality degradation of a decoded audio signal of each channel without requiring a delay or a computational processing amount for obtaining a monaural local decoded signal.

[0011] Problems to be solved by the invention

[0012] A sound signal encoding method according to one embodiment of the present invention encodes an input sound signal for each frame, and is characterized in that it includes: a downmixing step of obtaining a signal obtained by mixing an input left channel input sound signal and an input right channel input sound signal, namely a downmixed signal; a left channel subtraction gain estimation step of obtaining a left channel subtraction gain α and a code representing the left channel subtraction gain α, namely a left channel subtraction gain code Cα, based on the left channel input sound signal and the downmixed signal; a left channel signal subtraction step of obtaining a sample value x of the left channel input sound signal for each corresponding sample t. L (t) minus the sample value x of the downmixed signal M (t) is multiplied by the value of the left channel subtraction gain α to obtain the value x L (t)-α×x M (t) as the left channel differential signal; a right channel subtraction gain estimation step, according to the right channel input sound signal and the downmixed signal, obtaining the right channel subtraction gain β and the code representing the right channel subtraction gain β, i.e., the right channel subtraction gain code Cβ; a right channel signal subtraction step, for each corresponding sample t, obtaining the sample value x of the right channel input sound signal R (t) minus the sample value x of the downmixed signal M The value x obtained by multiplying (t) by the right channel subtraction gain β R (t)-β×x M (t) as the right channel differential signal; a mono encoding step, encoding the downmixed signal and obtaining a mono code CM; and a stereo encoding step, encoding the left channel differential signal and the right channel differential signal and obtaining a stereo code CS, and setting the number of bits used for encoding the downmixed signal in the mono encoding step to b M, the number of bits used for encoding the left channel difference signal in the stereo encoding step is set to b L , the number of bits used for encoding the right channel difference signal in the stereo encoding step is set to b R , in the left channel subtraction gain estimation step, the left channel correction coefficient c is obtained L The normalized inner product value r of the downmixed signal for the left channel input sound signal L The quantized value of the multiplication value is used as the left channel subtraction gain α, and the inner product value r with the left channel subtraction gain α or the normalized inner product value r is obtained. L The quantized value corresponding to the encoding is used as the left channel subtraction gain encoding Cα, the left channel correction coefficient c L is a value greater than 0 and less than 1, and is in b L =b M When 0.5, b L B M The more it is, the closer it is to 0 than 0.5. L B M The smaller the value, the closer it is to 1 than 0.5. In the right channel subtraction gain estimation step, the right channel correction coefficient c is obtained. R The normalized inner product value r of the downmixed signal for the right channel input sound signal R The quantized value of the multiplication value is used as the right channel subtraction gain β, and the inner product value r with the right channel subtraction gain β or the normalized inner product value r is obtained. R The quantized value corresponding to the encoding is used as the right channel subtraction gain encoding Cβ, the right channel correction coefficient c R is a value greater than 0 and less than 1, and is in b R =b M When 0.5, b R B M The more it is, the closer it is to 0 than 0.5. R B M The smaller the value, the closer it is to 1 than 0.5.

[0013] A sound signal encoding method according to one embodiment of the present invention encodes an input sound signal for each frame, and is characterized in that it includes: a downmixing step of obtaining a signal obtained by mixing an input left channel input sound signal and an input right channel input sound signal, namely a downmixed signal; a left channel subtraction gain estimation step of obtaining a left channel subtraction gain α and a code representing the left channel subtraction gain α, namely a left channel subtraction gain code Cα, based on the left channel input sound signal and the downmixed signal; and a left channel signal subtraction step of obtaining a sample value x of the left channel input sound signal for each corresponding sample t. L (t) minus the sample value x of the downmixed signal M (t) is multiplied by the value of the left channel subtraction gain α to obtain the value x L(t)-α×x M (t) as the left channel differential signal; a right channel subtraction gain estimation step, according to the right channel input sound signal and the downmixed signal, obtaining the right channel subtraction gain β and the code representing the right channel subtraction gain β, i.e., the right channel subtraction gain code Cβ; a right channel signal subtraction step, for each corresponding sample t, obtaining the sample value x of the right channel input sound signal R (t) minus the sample value x of the downmixed signal M The value x obtained by multiplying (t) by the right channel subtraction gain β R (t)-β×x M (t) as the right channel differential signal; a mono encoding step, encoding the downmixed signal and obtaining a mono code CM; and a stereo encoding step, encoding the left channel differential signal and the right channel differential signal and obtaining a stereo code CS, and setting the number of bits used for encoding the downmixed signal in the mono encoding step to b M , the number of bits used for encoding the left channel difference signal in the stereo encoding step is set to b L , the number of bits used for encoding the right channel difference signal in the stereo encoding step is set to b R , in the left channel subtraction gain estimation step, the left channel correction coefficient c is obtained L , the normalized inner product value r of the downmix signal for the left channel input sound signal L , and a pre-set value greater than 0 and less than 1, that is, a quantized value of the multiplication value of the left channel coefficient value as the left channel subtraction gain α, and obtain the normalized inner product value r with the left channel subtraction gain α L The quantized value, or the normalized inner product value r L The code corresponding to the quantized value of the value multiplied by the coefficient value of the left channel is used as the left channel subtraction gain code Cα, and the left channel correction coefficient c L is a value greater than 0 and less than 1, and is in b L =b M When 0.5, b L B M The more it is, the closer it is to 0 than 0.5. L B M The smaller the value, the closer it is to 1 than 0.5. In the right channel subtraction gain estimation step, the right channel correction coefficient c is obtained. R , the normalized inner product value r of the downmix signal for the right channel input sound signal R , and a pre-set value greater than 0 and less than 1, that is, a quantized value of the multiplication value of the right channel coefficient value as the right channel subtraction gain β, and obtain the normalized inner product value r with the right channel subtraction gain β R The quantized value or the normalized inner product value r RThe code corresponding to the quantized value of the value multiplied by the coefficient value of the right channel is used as the right channel subtraction gain code Cβ, and the right channel correction coefficient c R is a value greater than 0 and less than 1, and is in b R =b M When 0.5, b R B M The more it is, the closer it is to 0 than 0.5. R B M The smaller the value, the closer it is to 1 than 0.5.

[0014] A sound signal encoding method according to one embodiment of the present invention encodes an input sound signal for each frame, and is characterized in that it includes: a downmixing step of obtaining a signal obtained by mixing an input left channel input sound signal and an input right channel input sound signal, namely a downmixed signal; a left channel subtraction gain estimation step of obtaining a left channel subtraction gain α and a code representing the left channel subtraction gain α, namely a left channel subtraction gain code Cα, based on the left channel input sound signal and the downmixed signal; and a left channel signal subtraction step of obtaining a sample value x of the left channel input sound signal for each corresponding sample t. L (t) minus the sample value x of the downmixed signal M (t) is multiplied by the value of the left channel subtraction gain α to obtain the value x L (t)-α×x M (t) as the left channel differential signal; a right channel subtraction gain estimation step, according to the right channel input sound signal and the downmixed signal, obtaining the right channel subtraction gain β and the code representing the right channel subtraction gain β, i.e., the right channel subtraction gain code Cβ; a right channel signal subtraction step, for each corresponding sample t, obtaining the sample value x of the right channel input sound signal R (t) minus the sample value x of the downmixed signal M The value x obtained by multiplying (t) by the right channel subtraction gain β R (t)-β×x M (t) as the right channel differential signal; a mono encoding step, encoding the downmixed signal and obtaining a mono code CM; and a stereo encoding step, encoding the left channel differential signal and the right channel differential signal and obtaining a stereo code CS, and setting the number of bits used for encoding the downmixed signal in the mono encoding step to b M , the number of bits used for encoding the left channel difference signal in the stereo encoding step is set to b L , the number of bits used for encoding the right channel difference signal in the stereo encoding step is set to b R , in the left channel subtraction gain estimation step, the left channel correction coefficient c is obtained L, the normalized inner product value r of the downmix signal for the left channel input sound signal L , the value greater than 0 and less than 1 determined for each frame, that is, the quantized value of the multiplication value of the left channel coefficient value is used as the left channel subtraction gain α, and the normalized inner product value r with the left channel subtraction gain α is obtained L The quantized value, or the normalized inner product value r L The code corresponding to the quantized value of the value multiplied by the coefficient value of the left channel is used as the left channel subtraction gain code Cα, and the left channel correction coefficient c L is a value greater than 0 and less than 1, and is in b L =b M When 0.5, b L B M The more it is, the closer it is to 0 than 0.5. L B M The smaller the value, the closer it is to 1 than 0.5. In the right channel subtraction gain estimation step, the right channel correction coefficient c is obtained. R , the normalized inner product value r of the downmix signal for the right channel input sound signal R , the value greater than 0 and less than 1 determined for each frame, that is, the quantized value of the multiplication value of the coefficient value for the right channel is used as the right channel subtraction gain β, and the normalized inner product value r with the right channel subtraction gain β is obtained R The quantized value, or the normalized inner product value r R The code corresponding to the quantized value of the value multiplied by the coefficient value of the right channel is used as the right channel subtraction gain code Cβ, and the right channel correction coefficient c R is a value greater than 0 and less than 1, and is in b R =b M When 0.5, b R B M The more it is, the closer it is to 0 than 0.5. R B M The smaller the value, the closer it is to 1 than 0.5.

[0015] A sound signal decoding method according to one embodiment of the present invention decodes an input code for each frame and obtains a sound signal, and is characterized in that it includes: a mono decoding step, decoding an input mono code CM and obtaining a mono decoded sound signal; a stereo decoding step, decoding an input stereo code CS and obtaining a left channel decoded differential signal and a right channel decoded differential signal; a left channel subtraction gain decoding step, decoding an input left channel subtraction gain code Cα and obtaining a left channel subtraction gain α; a left channel signal addition step, adding a sample value of the left channel decoded differential signal to the left channel according to each corresponding sample t. L (t), and the sampling value of the mono decoded sound signal ^x MThe value obtained by adding (t) and the value obtained by multiplying the left channel subtraction gain α^y L (t)+α×^x M (t) sequence is obtained as the left channel decoded sound signal; the right channel subtraction gain decoding step, the input right channel subtraction gain code Cβ is decoded and the right channel subtraction gain β is obtained; the right channel signal addition step, according to each corresponding sample t, the sampling value ^y of the right channel decoded differential signal R (t), and the sampling value of the mono decoded sound signal ^x M The value obtained by adding (t) and the value obtained by multiplying the right channel subtraction gain β^y R (t)+β×^x M The sequence of (t) is obtained as the right channel decoded audio signal, and the number of bits used for decoding the mono decoded audio signal in the mono decoding step is set to b M , the number of bits used for decoding the left channel decoding differential signal in the stereo decoding step is set to b L , the number of bits used for decoding the right channel decoding differential signal in the stereo decoding step is set to b R In the left channel subtraction gain decoding step, the left channel subtraction gain code Cα is decoded and a decoded value ^r is obtained. L , get the left channel correction coefficient c L and the decoded value obtained by decoding the left channel subtraction gain code Cα^r L The multiplication value is used as the left channel subtraction gain α, and the left channel correction coefficient c L is a value greater than 0 and less than 1, and is in b L =b M When 0.5, b L B M The more it is, the closer it is to 0 than 0.5. L B M The smaller the value, the closer it is to 1 than 0.5. In the right channel subtraction gain decoding step, the right channel subtraction gain code Cβ is decoded and the decoded value ^r is obtained. R , get the right channel correction coefficient c R and the decoded value obtained by decoding the right channel subtraction gain code Cβ^r R The multiplication value is used as the right channel subtraction gain β, and the right channel correction coefficient c R is a value greater than 0 and less than 1, and is in b R =b M When 0.5, b R B M The more it is, the closer it is to 0 than 0.5. R B M The smaller the value, the closer it is to 1 than 0.5.

[0016] Effects of the Invention

[0017] According to the present invention, embedded coding / decoding is provided that can suppress the sound quality degradation of decoded audio signals of each channel for a two-channel audio signal without increasing the delay or the amount of computational processing for obtaining a monaural local decoded signal. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 This is a block diagram showing an example of the encoding device according to the first embodiment and the fourth embodiment.

[0019] Figure 2 This is a flowchart showing an example of processing by the encoding device according to the first embodiment.

[0020] Figure 3 This is a block diagram showing an example of a decoding device according to the first embodiment.

[0021] Figure 4 This is a flowchart showing an example of processing by the decoding device according to the first embodiment.

[0022] Figure 5 1 is a flowchart showing an example of processing by the left channel subtraction gain estimation unit and the right channel subtraction gain estimation unit according to the first embodiment.

[0023] Figure 6 1 is a flowchart showing an example of processing by the left channel subtraction gain estimation unit and the right channel subtraction gain estimation unit according to the first embodiment.

[0024] Figure 7 1 is a flowchart showing an example of processing by the left channel subtraction gain decoding unit and the right channel subtraction gain decoding unit according to the first embodiment.

[0025] Figure 8 1 is a flowchart showing an example of processing by the left channel subtraction gain estimation unit and the right channel subtraction gain estimation unit according to the first embodiment.

[0026] Fig. 9 1 is a flowchart showing an example of processing by the left channel subtraction gain estimation unit and the right channel subtraction gain estimation unit according to the first embodiment.

[0027] Fig.10 This is a block diagram showing an example of an encoding device according to the second embodiment and the third embodiment.

[0028] Fig.11 This is a flowchart showing an example of processing by the encoding device according to the second embodiment.

[0029] Fig.12 This is a block diagram showing an example of a decoding device according to the second embodiment.

[0030] Fig.13 This is a flowchart showing an example of processing by the decoding device according to the second embodiment.

[0031] Fig.14 This is a flowchart showing an example of processing by the encoding device according to the third embodiment.

[0032] Fig.15 This is a flowchart showing an example of processing by the encoding device according to the fourth embodiment.

[0033] Fig.16 This is a diagram showing an example of a functional configuration of a computer that realizes each device in the embodiment of the present invention. DETAILED DESCRIPTION

[0034] <First Embodiment>

[0035] The encoding device and decoding device of the first embodiment are described. In addition, in the specification and claims, the encoding device is sometimes referred to as a sound signal encoding device, the encoding method is sometimes referred to as a sound signal encoding method, the decoding device is sometimes referred to as a sound signal decoding device, and the decoding method is sometimes referred to as a sound signal decoding method.

[0036] <<Encoding device 100>>

[0037] like Figure 1 As shown, the encoding device 100 of the first embodiment includes a downmixing unit 110, a left channel subtraction gain estimation unit 120, a left channel signal subtraction unit 130, a right channel subtraction gain estimation unit 140, a right channel signal subtraction unit 150, a monaural encoding unit 160, and a stereo encoding unit 170. The encoding device 100 encodes the input two-channel stereo time domain sound signal in units of a predetermined time length of 20 ms, for example, to obtain and output the monaural code CM, the left channel subtraction gain code Cα, the right channel subtraction gain code Cβ, and the stereo code CS described later. The two-channel stereo time domain sound signal input to the encoding device is, for example, a digital sound signal or a sound signal obtained by picking up sounds such as voices and music with two microphones and performing AD conversion, and is composed of a left channel input sound signal and a right channel input sound signal. The codes outputted by the encoding device, namely the mono code CM, the left channel subtraction gain code Cα, the right channel subtraction gain code Cβ and the stereo code CS, are inputted into the decoding device. The encoding device 100 performs the following steps for each frame: Figure 2 The illustrated processing is from step S110 to step S170.

[0038] [Downmixing unit 110]

[0039] The input audio signal of the left channel input to the encoding device 100 and the input audio signal of the right channel input to the encoding device 100 are input to the downmixing unit 110. The downmixing unit 110 obtains a downmixed signal, which is a signal obtained by mixing the input audio signal of the left channel and the input audio signal of the right channel, from the input audio signal of the left channel and the input audio signal of the right channel, and outputs the downmixed signal (step S110).

[0040] For example, when the number of samples per frame is T, the input audio signal x of the left channel input to the encoding device 100 in units of frames is input to the downmixing unit 110. L (1),x L (2),...,x L (T) and the input sound signal x of the right channel R (1),x R (2),...,x R (T). Here, T is a positive integer. For example, when the frame length is 20 ms and the sampling frequency is 32 kHz, T is 640. The downmixing unit 110 uses a sequence of average values ​​of the sample values ​​of each corresponding sample of the input left channel input sound signal and the input right channel input sound signal as the downmixed signal x. M (1),x M (2),...,x M (T) is obtained and output. That is, if each sample number is t, then x M (t) = (x L (t)+x R (t)) / 2.

[0041] [Left Channel Subtraction Gain Estimation Unit 120]

[0042] The left channel subtraction gain estimation unit 120 receives the input audio signal x of the left channel input to the encoding device 100. L (1),x L (2),...,x L (T) and the downmix signal x output by the downmixing unit 110 M (1),x M (2),...,x M (T). The left channel subtraction gain estimation unit 120 obtains the left channel subtraction gain α and the code representing the left channel subtraction gain α, i.e., the left channel subtraction gain code Cα, based on the input sound signal of the left channel and the downmix signal, and outputs them (step S120). The left channel subtraction gain estimation unit 120 obtains the left channel subtraction gain α and the left channel subtraction gain code Cα by a method based on the principle of minimizing the quantization error. The principle of minimizing the quantization error and the method based on the principle will be described later.

[0043] [Left channel signal subtraction unit 130]

[0044] The left channel signal subtraction unit 130 receives the input audio signal x of the left channel input to the encoding device 100. L (1),x L (2),...,x L (T), the downmixed signal x output by the downmixing unit 110 M (1),x M (2),...,x M (T), the left channel subtraction gain α output by the left channel subtraction gain estimation unit 120. The left channel signal subtraction unit 130 calculates the sample value x of the input audio signal of the left channel for each corresponding sample t. L (t) Subtract the sample value x of the downmixed signal M (t) is multiplied by the left channel subtraction gain α M (t) and the value x L (t)-α×x M (t) as the left channel differential signal y L (1),y L (2),...,y L (T) obtains and outputs (step S130). L (t) = x L (t)-α×x M (t). In the conventional encoding apparatus such as the non-patent document 1, the left channel difference signal is obtained by using not the downmix signal but the monaural encoded local decoded signal, that is, the quantized downmix signal. However, in the encoding apparatus 100, in order to avoid the delay or the amount of calculation processing for obtaining the local decoded signal, the left channel signal subtraction unit 130 uses not the monaural encoded local decoded signal, that is, the quantized downmix signal, but the unquantized downmix signal x obtained by the downmix unit 110. M (t).

[0045] [Right Channel Subtraction Gain Estimation Unit 140]

[0046] The right channel subtraction gain estimation unit 140 receives the input audio signal x of the right channel input to the encoding device 100. R (1),x R (2),...,x R (T) and the downmix signal x output from the downmixing unit 110 M (1),x M (2),...,x M(T). The right channel subtraction gain estimation unit 140 obtains the right channel subtraction gain β and the code representing the right channel subtraction gain β, i.e., the right channel subtraction gain code Cβ, based on the input sound signal of the input right channel and the downmix signal, and outputs them (step S140). The right channel subtraction gain estimation unit 140 obtains the right channel subtraction gain β and the right channel subtraction gain code Cβ by a method based on the principle of minimizing the quantization error. The principle of minimizing the quantization error and the method based on the principle will be described later.

[0047] [Right channel signal subtraction unit 150]

[0048] The right channel signal subtraction unit 150 receives the right channel input audio signal x input to the encoding device 100. R (1),x R (2),...,x R (T), the downmixed signal x output by the downmixing unit 110 M (1),x M (2),...,x M (T) and the right channel subtraction gain β output by the right channel subtraction gain estimation unit 140. The right channel signal subtraction unit 150 calculates the sampling value x of the input audio signal of the right channel for each corresponding sample t. R (t) Subtract the sample value x of the downmixed signal M (t) is multiplied by the right channel subtraction gain β M (t) and the value x R (t)-β×x M (t) as the right channel differential signal y R (1),y R (2),...,y R (T) obtains and outputs (step S150). R (t) = x R (t)-β×x M In the right channel signal subtraction unit 150, similarly to the left channel signal subtraction unit 130, in order to avoid the delay and the amount of computation required to obtain the local decoded signal in the encoding device 100, the unquantized downmix signal x obtained by the downmix unit 110 is used instead of the quantized downmix signal of the local decoded signal encoded as the monaural signal. M (t).

[0049] [Monaural encoding unit 160]

[0050] The downmix signal x output by the downmixing unit 110 is input to the monaural encoding unit 160. M (1),x M (2),...,x M(T). The mono encoding unit 160 uses a predetermined encoding method to encode the M The input downmix signal is encoded by 10 bits to obtain a mono code CM and output it (step S160). M (1),x M (2),...,x M (T) get b M The CM is encoded into a monaural channel of bits and outputted. Any encoding method may be used as the encoding method, and an encoding method such as the 3GPP EVS standard may be used.

[0051] [Stereo encoding unit 170]

[0052] The left channel difference signal y output from the left channel signal subtraction unit 130 is input to the stereo encoding unit 170. L (1),y L (2),...,y L (T), and the right channel difference signal y output by the right channel signal subtraction unit 150 R (1),y R (2),...,y R (T). The stereo encoding unit 170 uses a predetermined encoding method to encode the total b s The input left channel difference signal and right channel difference signal are encoded by bit to obtain a stereo code CS and output (step S170). L (1),y L (2),...,y L (T) and the input T sampled right channel differential signal y R (1),y R (2),...,y R (T) Get the total b s As the encoding method, any method can be used, for example, a stereo encoding method corresponding to the stereo decoding method of the MPEG-4AAC standard can be used, or a method of encoding the input left channel differential signal and the right channel differential signal independently can be used, or the encoding obtained by combining all the encodings can be used as the stereo encoding CS.

[0053] When the input left channel difference signal and right channel difference signal are encoded independently, the stereo encoding unit 170 uses b L The left channel differential signal is encoded with b R That is, the stereo encoding unit 170 encodes the right channel difference signal y from the input T sampled left channel difference signal yL (1),y L (2),...,y L (T) get b L The left channel differential code CL of bits is obtained from the right channel differential signal y sampled from the input T R (1),y R (2),...,y R (T) get b R The right channel differential code CR of b bits is combined with the left channel differential code CL and the right channel differential code CR to output as a stereo code CS. L Bit and b R The total number of bits is b S Bit.

[0054] When the input left channel difference signal and right channel difference signal are encoded together in one encoding method, the stereo encoding unit 170 uses the total b S That is, the stereo encoding unit 170 encodes the left channel difference signal y of the input T samples. L (1),y L (2),...,y L (T) and the input T sampled right channel differential signal y R (1),y R (2),...,y R (T) get b S The stereo code CS of the bits is output.

[0055] <<Decoding device 200>>

[0056] like Figure 3 As shown, the decoding device 200 of the first embodiment includes: a mono decoding unit 210, a stereo decoding unit 220, a left channel subtraction gain decoding unit 230, a left channel signal adding unit 240, a right channel subtraction gain decoding unit 250 and a right channel signal adding unit 260. The decoding device 200 decodes the input mono code CM and the left channel subtraction gain code Cα, the right channel subtraction gain code Cβ and the stereo code CS in frame units of the same time length as the corresponding encoding device 100, obtains and outputs a two-channel stereo time domain decoded sound signal (a left channel decoded sound signal and a right channel decoded sound signal to be described later) in frame units. The decoding device 200 may also be as follows Figure 3As shown by the dotted line in the middle, a decoded audio signal in the time domain of a monophonic sound is output (a monophonic decoded audio signal described later). For example, the decoded audio signal output by the decoding device 200 is reproduced by a speaker by performing DA conversion, so that the decoded audio signal can be received. The decoding device 200 performs Figure 4 The processing of steps S210 to S260 illustrated in FIG.

[0057] [Monaural decoding unit 210]

[0058] The monaural code CM input to the decoding device 200 is input to the monaural decoding unit 210. The monaural decoding unit 210 decodes the input monaural code CM using a predetermined decoding method to obtain a monaural decoded audio signal. M (1),^x M (2),...,^x M (T) and output (step S210). As the predetermined decoding method, a decoding method corresponding to the encoding method used in the monaural encoding unit 160 of the corresponding encoding device 100 is used. The number of bits of the monaural code CM is b M .

[0059] [Stereo decoding unit 220]

[0060] The stereo code CS input to the decoding device 200 is input to the stereo decoding unit 220. The stereo decoding unit 220 decodes the input stereo code CS using a predetermined decoding method to obtain a left channel decoded differential signal y. L (1),^y L (2),...,^y L (T), right channel decoded differential signal ^y R (1),^y R (2),...,^y R (T) and outputs (step S220). As the predetermined decoding method, a decoding method corresponding to the encoding method used in the stereo encoding unit 170 of the corresponding encoding device 100 is used. The total number of bits of the stereo code CS is b S .

[0061] [Left channel subtraction gain decoding unit 230]

[0062] The left channel subtraction gain code Cα input to the decoding device 200 is input to the left channel subtraction gain decoding unit 230. The left channel subtraction gain decoding unit 230 decodes the left channel subtraction gain code Cα and obtains the left channel subtraction gain α and outputs it (step S230). The method by which the left channel subtraction gain decoding unit 230 decodes the left channel subtraction gain code Cα and obtains the left channel subtraction gain α will be described later.

[0063] [Left channel signal adding section 240]

[0064] The monaural decoded audio signal outputted from the monaural decoding unit 210 is inputted to the left channel signal adding unit 240. M (1),^x M (2),...,^x M (T), the left channel decoded differential signal output by the stereo decoding unit 220 L (1),^y L (2),...,^y L (T), the left channel subtraction gain α output by the left channel subtraction gain decoding unit 230. The left channel signal adding unit 240 converts the sample value of the left channel decoded difference signal t into a value y for each corresponding sample t. L (t), and the sampling value of the mono decoded sound signal ^x M (t) is multiplied by the left channel subtraction gain α×^x M (t) The added value ^y L (t)+α×^x M The sequence of (t) is used as the left channel decoded sound signal ^x L (1),^x L (2),...,^x L (T) is obtained and output (step S240). L (t) = ^y L (t)+α×^x M (t).

[0065] [Right Channel Subtraction Gain Decoding Unit 250]

[0066] The right channel subtraction gain code Cβ input to the decoding device 200 is input to the right channel subtraction gain decoding unit 250. The right channel subtraction gain decoding unit 250 decodes the right channel subtraction gain code Cβ and obtains the right channel subtraction gain β and outputs it (step S250). The method by which the right channel subtraction gain decoding unit 250 decodes the right channel subtraction gain code Cβ and obtains the right channel subtraction gain β will be described later.

[0067] [Right channel signal adding unit 260]

[0068] The right channel signal adding unit 260 receives the monaural decoded audio signal ^x outputted from the monaural decoding unit 210. M (1),^x M (2),...,^x M (T), the right channel decoded differential signal output by the stereo decoding unit 220 R (1),^yR (2),...,^y R (T), the right channel subtraction gain β output by the right channel subtraction gain decoding unit 250. The right channel signal adding unit 260 converts the sample value of the right channel decoded difference signal t into a value y for each corresponding sample t. R (t), and the sampling value of the mono decoded sound signal ^x M (t) multiplied by the right channel subtraction gain β M (t) The value obtained by adding ^y R (t)+β×^x M The sequence of (t) is used as the right channel decoded sound signal ^x R (1),^x R (2),...,^x R (T) is performed and output (step S260). R (t) = ^y R (t)+β×^x M (t).

[0069] [Principle of minimizing quantization error]

[0070] The principle of minimizing the quantization error is described below. When the left channel difference signal and the right channel difference signal input to the stereo encoding unit 170 are combined and encoded in one encoding method, the number of bits b used in encoding the left channel difference signal is L The number of bits used in encoding the right channel difference signal is b R It may not be clearly determined, but the following description assumes that the number of bits used in encoding the left channel differential signal is b. L , the number of bits used in encoding the right channel differential signal is b R In addition, the following description mainly focuses on the left channel, but the same is true for the right channel.

[0071] The above-mentioned encoding device 100 uses b L The bit pair is the input audio signal x from the left channel L (1),x L (2),...,x L Each sample value of (T) is subtracted from the downmix signal x M (1),x M (2),...,x M The left channel difference signal y is composed of the value obtained by multiplying each sample value of (T) by the value of the left channel subtraction gain α L (1),y L (2),...,y L (T) is encoded and b is used M Bit-pair downmix signal xM (1),x M (2),...,x M (T) is encoded. In addition, the decoding device 200 is based on b L Bit encoding for left channel decoded differential signal ^y L (1),^y L (2),...,^y L (hereinafter also referred to as "quantized left channel differential signal") is decoded according to b M Bit encoding for mono decoded audio signal^x M (1),^x M (2),...,^x M After decoding (T) (hereinafter also referred to as "quantized downmix signal"), the quantized downmix signal obtained by decoding is converted into M (1),^x M (2),...,^x M The value obtained by multiplying each sample value of (T) by the left channel gain α is added to the quantized left channel difference signal ^y obtained by decoding L (1),^y L (2),...,^y L (T) each sampling value, thereby obtaining the left channel decoded audio signal ^x as the left channel decoded audio signal L (1),^x L (2),...,^x L (T) The encoding apparatus 100 and the decoding apparatus 200 should be designed to reduce the energy of the quantization error of the decoded audio signal of the left channel obtained in the above-mentioned process.

[0072] The energy of the quantization error (hereinafter, for convenience, referred to as "quantization error caused by encoding") in the decoded signal obtained by encoding / decoding the input signal is generally proportional to the energy of the input signal, and tends to decrease exponentially with respect to the value of the number of bits per sample used in the encoding. Therefore, the average energy per sample of the quantization error caused by encoding the left channel difference signal is expressed as a positive number σ L 2 It can be estimated as shown in the following equation (1-0-1). The average energy per sample of the quantization error generated by encoding the downmix signal is expressed as a positive number σ M 2 It can be estimated as shown in the following formula (1-0-2).

[0073] [Mathematical formula 1]

[0074]

[0075] [Mathematical formula 2]

[0076]

[0077] Here, let the input audio signal x of the left channel be L (1),x L (2),...,x L (T) and the downmixed signal x M (1),x M (2),...,x M (T) is the value of each sample value that is approximately regarded as the same sequence. For example, the input sound signal x of the left channel L (1),x L (2),...,x L (T) and the input signal x of the right channel R (1),x R (2),...,x R (T) is equivalent to the case where the sound emitted by the sound source at an equal distance from the two microphones is collected in an environment with a lot of background noise or reverberation. Under this condition, the left channel differential signal y L (1),y L (2),...,y L The sampling values ​​of (T) and the downmixed signal x M (1),x M (2),...,x M The value obtained by multiplying each sample value of (T) by (1-α) is equivalent. Therefore, the energy of the left channel difference signal is (1-α) times the energy of the downmix signal. 2 times, so the above σ L 2 It is possible to use the above σ M 2 Replace with (1-α) 2 ×σ M 2 Therefore, the average energy per sample of the quantization error generated by encoding the left channel difference signal can be estimated as shown in the following formula (1-1).

[0078] [Mathematical formula 3]

[0079]

[0080] In addition, the average energy per sample of the quantization error of the signal to which the quantized left channel differential signal is added in the decoding device, that is, the average energy per sample of the quantization error of the sequence of values ​​obtained by multiplying each sample value of the quantized downmix signal obtained by decoding and the left channel subtraction gain α, can be estimated as shown in the following formula (1-2).

[0081] [Formula 4]

[0082]

[0083] If it is assumed that the quantization error generated by encoding the left channel difference signal and the quantization error of the sequence of values ​​obtained by multiplying each sample value of the quantized downmix signal obtained by decoding by the left channel subtraction gain α have no correlation with each other, the average energy per sample of the quantization error of the decoded sound signal of the left channel is estimated by the sum of equations (1-1) and (1-2). The left channel subtraction gain α that minimizes the energy of the quantization error of the decoded sound signal of the left channel is obtained as shown in the following equation (1-3).

[0084] [Formula 5]

[0085]

[0086] That is, the input sound signal x in the left channel L (1),x L (2),...,x L (T) and downmix signal xx M (1),x M (2),...,x M (T) Under the condition that the values ​​of each sample value are approximately regarded as the same sequence, in order to minimize the quantization error of the decoded sound signal of the left channel, the left channel subtraction gain estimation unit 120 can obtain the left channel subtraction gain α through equation (1-3). The left channel subtraction gain α obtained by equation (1-3) is a value greater than 0 and less than 1. When the number of bits used for 2 encodings is b L With b M When they are equal, it is 0.5, which is the number of bits b used to encode the left channel differential signal. L is greater than the number of bits b used to encode the downmix signal M The closer the value is to 0.5, the more bits b are used to encode the downmix signal. M Greater than the number of bits b used to encode the left channel differential signal L The closer it is to 0.5.

[0087] The same is true for the right channel. The input sound signal x R(1),x R (2),...,x R (T), downmix signal x M (1),x M (2),...,x M Under the condition that the values ​​of each sample value (T) are approximately regarded as the same sequence, in order to minimize the quantization error of the decoded audio signal of the right channel, the right channel subtraction gain estimation unit 140 can calculate the right channel subtraction gain β by the following formula (1-3-2).

[0088] [Mathematical formula 6]

[0089]

[0090] The right channel subtraction gain β obtained by formula (1-3-2) is a value greater than 0 and less than 1. The number of bits used in the two codes is b R With b M When they are equal, it is 0.5, which is the number of bits b used to encode the right channel differential signal R The more bits b are used to encode the downmix signal M The closer the value is to 0.5, the number of bits b used to encode the downmix signal M The more bits b are used to encode the right channel differential signal R The more it is, the closer it is to 0.5.

[0091] Next, the input audio signal x including the left channel L (1),x L (2),...,x L (T) and the downmixed signal x M (1),x M (2),...,x M (T) The principle of minimizing the energy of the quantization error possessed by the decoded audio signal of the left channel, including the case where the decoded audio signal cannot be regarded as the same sequence, is explained.

[0092] The input sound signal of the left channel is x L (1),x L (2),...,x L (T) and the downmixed signal x M (1),x M (2),...,x M The normalized inner product value r of (T) L It is expressed by the following formula (1-4).

[0093] [Formula 7]

[0094]

[0095] The normalized inner product value r obtained by formula (1-4) L is a real value that is related to the downmix signal x M (1),x M (2),...,x M Each sample value of (T) is multiplied by the real value r L 'And get the sequence r of sample values L '×x M (1),r L '×x M (2),...,r L '×x M The sequence x obtained by the difference between the sequence of sample values ​​obtained at (T) and each sample value of the input audio signal of the left channel is L (1)-r L '×x M (1),x L (2)-r L '×x M (2),...,x L (T)-r L '×x M The energy of (T) becomes the smallest real value r L Same value.

[0096] The input sound signal of the left channel is x L (1),x L (2),...,x L (T) For each sampling number t, it can be decomposed into x L (t) = r L × M (t)+(x L (t)-r L × M (t)). Here, when x L (t)-r L × M The sequence of values ​​of (t) is set as the orthogonal signal x L '(1),x L '(2),...,x L '(T), according to the decomposition, the sampling values ​​y of the left channel differential signal L (t) = x L (t)-αx M (t) and the downmixed signal x M (1),x M (2),...,x M Each sample value x of (T) M (t) is multiplied by the normalized inner product value rL and using the left channel subtraction gain α (r L -α) and the value (r L -α)×x M (t) and the sampling values ​​x of the orthogonal signal L '(t) and (r L -α)×x M (t)+x L '(t) is equivalent. Orthogonal signal x L '(1),x L '(2),...,x L '(T) represents the downmix signal x M (1),x M (2),...,x M (T) is orthogonal, that is, the property that the inner product is 0, so the energy of the left channel difference signal is obtained by multiplying the energy of the downmix signal by (r L -α) 2 Therefore, the energy of the quadrature signal is expressed as the sum of the energy of the quadrature signal and the energy of the orthogonal signal. L The average energy per sample of the quantization error generated by encoding the left channel difference signal is a positive number σ 2 , can be estimated as shown in the following formula (1-5).

[0097] [Mathematical formula 8]

[0098]

[0099] If it is assumed that the quantization error generated by encoding the left channel difference signal and the quantization error of the sequence of values ​​obtained by multiplying each sample value of the quantized downmix signal obtained by decoding by the left channel subtraction gain α have no correlation with each other, the average energy per sample of the quantization error of the decoded sound signal of the left channel is estimated by the sum of equation (1-5) and equation (1-2). The left channel subtraction gain α that minimizes the energy of the quantization error of the decoded sound signal of the left channel is obtained as shown in the following equation (1-6).

[0100] [Mathematical formula 9]

[0101]

[0102] That is, in order to minimize the quantization error of the decoded audio signal of the left channel, the left channel subtraction gain estimation unit 120 only needs to obtain the left channel subtraction gain α by using equation (1-6). In other words, if the principle of minimizing the energy of the quantization error is considered, the left channel subtraction gain α should be calculated by normalizing the inner product value r LThe value obtained by multiplying the correction factor by the number of bits used for encoding, i.e., b L and b M The correction coefficient is a value greater than 0 and less than 1, and is used to encode the left channel differential signal. L and the number of bits b used to encode the downmix signal M If the same, it is 0.5, which is the number of bits b used to encode the left channel differential signal. L The number of bits b used to encode the downmix signal M The more it is, the closer it is to 0.5, the number of bits b used to encode the left channel differential signal L The number of bits b used to encode the downmix signal M The smaller the value, the closer it is to 0.5.

[0103] The same is true for the right channel. In order to minimize the quantization error of the decoded audio signal of the right channel, the right channel subtraction gain estimation unit 140 may calculate the right channel subtraction gain β using the following equation (1-6-2).

[0104] [Formula 10]

[0105]

[0106] Here, r R is the input sound signal x of the right channel R (1),x R (2),...,x R (T) and the downmixed signal x M (1),x M (2),...,x M The normalized inner product value of (T) is expressed by the following formula (1-4-2).

[0107] [Mathematical formula 11]

[0108]

[0109] That is, if the principle of minimizing the energy of the quantization error is considered, the right channel subtraction gain β should use the normalized inner product value r R The value obtained by multiplying the correction factor by the number of bits used for encoding, i.e., b R and b M The correction coefficient is a value greater than 0 and less than 1, and is used to encode the number of bits b of the right channel differential signal. R The number of bits b used to encode the downmix signal MThe larger the value, the closer it is to 0.5. The smaller the number of bits used to encode the right channel difference signal is than the number of bits used to encode the downmix signal, the closer it is to 0.5.

[0110] [Estimation and decoding of subtraction gain based on the principle of minimizing quantization error]

[0111] Specific examples of estimation and decoding of subtraction gain based on the principle of minimizing the above-mentioned quantization error are described. In each example, the left channel subtraction gain estimation unit 120 and the right channel subtraction gain estimation unit 140 that estimate the subtraction gain in the encoding device 100, and the left channel subtraction gain decoding unit 230 and the right channel subtraction gain decoding unit 250 that decode the subtraction gain in the decoding device 200 are described.

[0112] [[Example 1]]

[0113] Example 1 is based on the following principle: including the input sound signal x of the left channel L (1),x L (2),...,x L (T) and the downmixed signal x M (1),x M (2),...,x M (T) The principle of minimizing the energy of the quantization error of the decoded audio signal of the left channel, including the case where the decoded audio signal of the left channel is not considered as the same series; and the principle of minimizing the energy of the quantization error of the decoded audio signal of the right channel R (1),x R (2),...,x R (T) and the downmixed signal x M (1),x M (2),...,x M (T) The principle of minimizing the energy of the quantization error of the decoded audio signal of the right channel, regardless of the case of not considering the same series.

[0114] [Left Channel Subtraction Gain Estimation Unit 120]

[0115] In the left channel subtraction gain estimation unit 120, the left channel subtraction gain candidate α cand (a) and the code Cα corresponding to the candidate cand The group (a) is stored in a plurality of groups (A groups, a=1, ..., A) in advance. The left channel subtraction gain estimation unit 120 performs the following operation: Figure 5 The following steps S120-11 to S120-14 are shown.

[0116] The left channel subtraction gain estimation unit 120 first calculates the left channel subtraction gain based on the input audio signal x L (1),x L(2),...,x L (T) and the downmixed signal x M (1),x M (2),...,x M (T), the normalized inner product value r of the input audio signal of the left channel of the downmix signal is obtained by equation (1-4): L (Step S120-11). Then, the left channel subtraction gain estimation unit 120 uses the left channel difference signal y in the stereo encoding unit 170 L (1),y L (2),...,y L (T) is the number of bits b L , used in the monaural encoding unit 160 to downmix the signal x M (1),x M (2),...,x M (T) is the number of bits b M , and the number of samples per frame T, the left channel correction coefficient c is obtained by the following formula (1-7): L (Step S120-12).

[0117] [Mathematical formula 12]

[0118]

[0119] Next, the left channel subtraction gain estimation unit 120 obtains the normalized inner product value r obtained in step S120-11. L The left channel correction coefficient c obtained in step S120-12 L Then, the left channel subtraction gain estimation unit 120 obtains the stored candidate α of the left channel subtraction gain. cand (1),...,α cand (A) is closest to the multiplication value c obtained in step S120-13 L × L Candidate (multiplication value c L × L The quantized value of the left channel is used as the left channel subtraction gain α to obtain the stored code Cα cand (1),...,Cα cand The code corresponding to the left channel subtraction gain α in (A) is taken as the left channel subtraction gain code Cα (step S120 - 14 ).

[0120] In addition, in the stereo encoding unit 170, the left channel difference signal y L (1),y L (2),...,y LThe number of bits b used in the encoding of (T) L If it is not clearly determined, the number of bits b of the stereo code CS output by the stereo coding unit 170 is s One-half of (i.e., b s / 2) is used as the bit number b L In addition, the left channel correction coefficient c L It may also be a value other than the value obtained by equation (1-7) itself, but a value greater than 0 and less than 1, and may be the following value: L (1),y L (2),...,y L (T) is the number of bits b L and for the downmix signal x M (1),x M (2),...,x M (T) is the number of bits b M If the same, it is 0.5, the number of bits is b L Bit number b M The more it is, the closer it is to 0 than 0.5. The number of bits b L Bit number b M The smaller the value, the closer it is to 1 than 0.5. This also applies to each example described below.

[0121] [Right Channel Subtraction Gain Estimation Unit 140]

[0122] In the right channel subtraction gain estimation unit 140, a plurality of sets (B sets, b=1, . . . , B) of candidate β values ​​for the right channel subtraction gain are stored in advance. cand (b) and the encoding Cβ corresponding to the candidate cand (b) group. The right channel subtraction gain estimation unit 140 performs the following Figure 5 The following steps S140-11 to S140-14 are shown.

[0123] The right channel subtraction gain estimation unit 140 first calculates the right channel subtraction gain based on the input audio signal x R (1),x R (2),...,x R (T) and the downmixed signal x M (1),x M (2),...,x M (T), the normalized inner product value r of the input audio signal of the right channel of the downmix signal is obtained by formula (1-4-2): R (Step S140-11). Then, the right channel subtraction gain estimation unit 140 uses the right channel difference signal y in the stereo encoding unit 170 R (1),yR (2),...,y R (T) is the number of bits b R , used in the monaural encoding unit 160 to downmix the signal x M (1),x M (2),...,x M (T) is the number of bits b M , and the number of samples per frame T, the right channel correction coefficient c is obtained by the following formula (1-7-2): R (Step S140-12).

[0124] [Mathematical formula 13]

[0125]

[0126] Next, the right channel subtraction gain estimation unit 140 obtains the normalized inner product value r obtained in step S140-11. R The right channel correction coefficient c obtained in step S140-12 R Next, the right channel subtraction gain estimation unit 140 obtains the candidate β closest to the stored right channel subtraction gain. cand (1),...,β cand The multiplication value c obtained in step S140-13 in (B) R × R Candidate (multiplication value c R × R The quantized value of the right channel is used as the subtraction gain β of the right channel to obtain the stored code Cβ cand (1),...,Cβ cand The code corresponding to the right channel subtraction gain β in (B) is used as the right channel subtraction gain Cβ (step S140-14).

[0127] In addition, in the stereo encoding unit 170, the right channel difference signal y R (1),y R (2),...,y R The number of bits b used in the encoding of (T) R If it is not clearly determined, the number of bits b of the stereo code CS output by the stereo coding unit 170 is s One-half of (i.e., b s / 2) is used as the bit number b R In addition, the right channel correction coefficient c R It can also be a value other than the value obtained by equation (1-7-2) itself, but a value greater than 0 and less than 1, and it can be the following value: R (1),yR (2),...,y R (T) is the number of bits b R and for the downmix signal x M (1),x M (2),...,x M (T)(T) is the number of bits b encoded M If the same, it is 0.5, the number of bits is b R Bit number b M The more it is, the closer it is to 0 than 0.5. The number of bits b R Bit number b M The smaller the value, the closer it is to 1 than 0.5. This also applies to each example described below.

[0128] [Left channel subtraction gain decoding unit 230]

[0129] In the left channel subtraction gain decoding unit 230, a plurality of sets (A sets, a=1, ..., A) of candidate α values ​​of the left channel subtraction gain are stored in advance, similarly to the portion stored in the corresponding left channel subtraction gain estimation unit 120 of the encoding device 100. cand (a) and the encoding Cα corresponding to the candidate cand The left channel subtraction gain decoding unit 230 converts the stored code Cα cand (1),...,Cα cand The candidate of the left channel subtraction gain corresponding to the input left channel subtraction gain code Cα in (A) is obtained as the left channel subtraction gain α (step S230 - 11 ).

[0130] [Right channel subtraction gain decoding unit 250]

[0131] In the right channel subtraction gain decoding unit 250, a plurality of sets (B sets, b=1, . . . , B) of candidate β values ​​of the right channel subtraction gain are pre-stored, similarly to the portion stored in the corresponding right channel subtraction gain estimation unit 140 of the encoding device 100. cand (b) and the encoding Cβ corresponding to the candidate cand (b) The right channel subtraction gain decoding unit 250 converts the stored code Cβ cand (1),...,Cβ cand The candidate of the right channel subtraction gain corresponding to the input right channel subtraction gain code Cβ in (B) is obtained as the right channel subtraction gain β (step S250 - 11 ).

[0132] Alternatively, the same subtraction gain candidate may be used for the left channel and the right channel, and the above-mentioned A and B may be set to the same value. Alternatively, the left channel subtraction gain candidate α stored in the left channel subtraction gain estimation unit 120 and the left channel subtraction gain decoding unit 230 may be set to α. cand (a) and the encoding Cα corresponding to the candidate cand (a), the candidate β of the right channel subtraction gain stored in the right channel subtraction gain estimation unit 140 and the right channel subtraction gain decoding unit 250 cand (b) and the encoding Cβ corresponding to the candidate cand The groups of (b) are the same.

[0133] [[Variation of Example 1]]

[0134] The number of bits b used for encoding the left channel differential signal in the encoding device 100 is L is the number of bits used for decoding the left channel difference signal in the decoding device 200, and is the number of bits used for encoding the downmix signal in the encoding device 100. M The value of is the number of bits used for decoding the downmix signal in the decoding apparatus 200, so the correction coefficient c L The same value can be calculated in the encoding device 100 and the decoding device 200. Therefore, the normalized inner product value r L As the object of encoding and decoding, the quantized value of the normalized inner product value is quantized in the encoding device 100 and the decoding device 200. L Multiply by the correction factor c L , the left channel subtraction gain α is obtained. The same is true for the right channel. This method is described as a modified example of Example 1.

[0135] [Left Channel Subtraction Gain Estimation Unit 120]

[0136] In the left channel subtraction gain estimation unit 120, a plurality of groups (A groups, a=1, . . . , A) of candidates for normalized inner product values ​​of the left channel are stored in advance. Lcand (a) and the encoding Cα corresponding to the candidate cand (a) group. Figure 6 As shown, the left channel subtraction gain estimation unit 120 performs steps S120-11 and S120-12 described in Example 1, and steps S120-15 and S120-16 described below.

[0137] The left channel subtraction gain estimation unit 120 first performs the same operation as step S120-11 of the left channel subtraction gain estimation unit 120 in Example 1, based on the input audio signal x of the left channel. L (1),x L (2),...,x L(T) and the downmixed signal x M (1),x M (2),...,x M (T), the normalized inner product value r of the input audio signal of the left channel of the downmix signal is obtained by equation (1-4): L (Step S120 - 11 ) Next, the left channel subtraction gain estimation unit 120 obtains a candidate r of the normalized inner product value with the stored left channel. Lcand (1),...,r Lcand The normalized inner product value r obtained by step S120-11 in (A) L The closest candidate (normalized inner product value r L Quantitative value of L , and get the stored code Cα cand (1),...,Cα cand (A) The closest candidate ^r L The corresponding code is used as the left channel subtraction gain code Cα (step S120-15). In addition, the left channel subtraction gain estimation unit 120 uses the left channel difference signal y in the stereo encoding unit 170 in the same manner as step S120-12 of the left channel subtraction gain estimation unit 120 in Example 1. L (1),y L (2),...,y L (T) is the number of bits b L , used in the monaural encoding unit 160 to downmix the signal x M (1),x M (2),...,x M (T) is the number of bits b M , and the number of samples per frame T, the left channel correction coefficient c is obtained by formula (1-7) L (Step S120-12) Next, the left channel subtraction gain estimation unit 120 obtains a quantized value of the normalized inner product value obtained in step S120-15. L The left channel correction coefficient c obtained in step S120-12 L The multiplied value is used as the left channel subtraction gain α (step S120-16).

[0138] [Right Channel Subtraction Gain Estimation Unit 140]

[0139] In the right channel subtraction gain estimation unit 140, a plurality of sets (B sets, b=1, . . . , B) of candidates for normalized inner product values ​​of the right channel are stored in advance. Rcand (b) and the encoding Cβ corresponding to the candidate cand (b) group. Figure 6As shown, the right channel subtraction gain estimation unit 140 performs steps S140-11 and S140-12 described in Example 1, and steps S140-15 and S140-16 described below.

[0140] The right channel subtraction gain estimation unit 140 first performs the same operation as step S140-11 of the right channel subtraction gain estimation unit 140 in Example 1, based on the input audio signal x of the right channel. R (1),x R (2),...,x R (T) and the downmixed signal x M (1),x M (2),...,x M (T), the normalized inner product value r of the input audio signal of the right channel of the downmix signal is obtained by formula (1-4-2): R (Step S140 - 11 ) Next, the right channel subtraction gain estimation unit 140 obtains a candidate r of the normalized inner product value with the stored right channel. Rcand (1),...,r Rcand (B) The normalized inner product value r obtained by step S140-11 R The closest candidate (normalized inner product value r R Quantitative value of R , and get the stored code Cβ cand (1),...,Cβ cand The closest candidate r in (B) R The corresponding code is used as the right channel subtraction gain code Cβ (step S140-15). Then, the right channel subtraction gain estimation unit 140 uses the right channel difference signal y in the stereo encoding unit 170 in the same manner as step S140-12 of the right channel subtraction gain estimation unit 140 in Example 1. R (1),y R (2),...,y R (T) is the number of bits b R , used in the monaural encoding unit 160 to downmix the signal x M (1),x M (2),...,x M (T) is the number of bits b M , and the number of samples per frame T, the right channel correction coefficient c is obtained by formula (1-7-2) R (Step S140-12) Next, the right channel subtraction gain estimation unit 140 obtains a quantized value of the normalized inner product value obtained in step S140-15. R The right channel correction coefficient c obtained in step S140-12 RThe multiplied value is used as the right channel subtraction gain β (step S140-16).

[0141] [Left channel subtraction gain decoding unit 230]

[0142] In the left channel subtraction gain decoding unit 230, similarly to the part stored in the corresponding left channel subtraction gain estimation unit 120 of the encoding device 100, a plurality of groups (A groups, a=1, ..., A) of candidates for normalized inner product values ​​of the left channel are pre-stored. Lcand (a) and the encoding Cα corresponding to the candidate cand The left channel subtraction gain decoding unit 230 performs the following operation: Figure 7 The following steps S230-12 to S230-14 are shown.

[0143] The left channel subtraction gain decoding unit 230 converts the stored code Cα cand (1),...,Cα cand The candidate of the normalized inner product value of the left channel corresponding to the input left channel subtraction gain code Cα in (A) obtains the decoded value ^r as the normalized inner product value of the left channel L (Step S230-12). In addition, the left channel subtraction gain decoding unit 230 uses the left channel decoded differential signal y in the stereo decoding unit 220. L (1),^y L (2),...,^y L (T) is the number of bits decoded L , used in the monaural decoding unit 210 to decode the audio signal in monaural mode M (1),^x M (2),...,^x M (T) is the number of bits decoded M , and the number of samples per frame T, the left channel correction coefficient c is obtained by formula (1-7) L (Step S230-13) Next, the left channel subtraction gain decoding unit 230 obtains a decoded value of the normalized inner product value obtained in step S230-12. L The left channel correction coefficient c obtained in step S230-13 L The multiplied value is used as the left channel subtraction gain α (step S230-14).

[0144] In addition, when the stereo code CS is a code obtained by combining the left channel difference code CL and the right channel difference code CR, the left channel decoded difference signal y is used in the stereo decoding unit 220. L (1),^y L (2),...,^yL (T) is the number of bits decoded L is the number of bits of the left channel differential code CL. L (1),^y L (2),...,^y L (T) The number of bits decoded L If it is not clearly determined, the number of bits b of the stereo code CS input to the stereo decoding unit 220 is s One-half of (i.e., b s / 2) is used as the bit number b L In the mono decoding unit 210, the mono decoding audio signal is decoded into M (1),^x M (2),...,^x M (T) is the number of bits decoded M is the number of bits of mono coded CM. Left channel correction coefficient c L It is not a value obtained by equation (1-7) itself, but a value greater than 0 and less than 1, and is the following value: L (1),^y L (2),...,^y L (T) The number of bits decoded L and for mono decoded sound signal^x M (1),^x M (2),...,^x M (T) is the number of bits decoded M If the same, it is 0.5, the number of bits is b L Bit number b M The more it is, the closer it is to 0 than 0.5. The number of bits b L Bit number b M The smaller it is, the closer it is to 1 than 0.5.

[0145] [Right channel subtraction gain decoding unit 250]

[0146] In the right channel subtraction gain decoding unit 250, similarly to the portion stored in the corresponding right channel subtraction gain estimation unit 140 of the encoding device 100, a plurality of sets (B sets, b=1, ..., B) of candidates for normalized inner product values ​​of the right channel are pre-stored. Rcand (b) and the encoding Cβ corresponding to the candidate cand (b) group. The right channel subtraction gain decoding unit 250 performs the following Figure 7 The following steps S250-12 to S250-14 are shown.

[0147] The right channel subtraction gain decoding unit 250 converts the stored code Cβ cand (1),...,Cβ cand The candidate of the normalized inner product value of the right channel corresponding to the input right channel subtraction gain code Cβ in (B) obtains the decoded value ^r as the normalized inner product value of the right channel R (Step S250-12). In addition, the right channel subtraction gain decoding unit 250 uses the differential signal φy used for right channel decoding in the stereo decoding unit 220. R (1),^y R (2),...,^y R (T) is the number of bits decoded R , in the mono decoding unit 210 for mono decoding the sound signal ^x M (1),^x M (2),...,^x M (T) is the number of bits decoded M , and the number of samples per frame T, the right channel correction coefficient c is obtained by formula (1-7-2) R (Step S250-13) Next, the right channel subtraction gain decoding unit 250 obtains a decoded value of the normalized inner product value obtained in step S250-12. R The right channel correction coefficient c obtained in step S250-13 R The value obtained by multiplying R is used as the right channel subtraction gain β (step S250-14).

[0148] In addition, when the stereo code CS is a code obtained by combining the left channel difference code CL and the right channel difference code CR, the right channel decoding difference signal y is used in the stereo decoding unit 220. R (1),^y R (2),...,^y R (T) is the number of bits decoded R is the number of bits of the right channel differential code CR. R (1),^y R (2),...,^y R (T) is the number of bits decoded R If it is not clearly determined, the number of bits b of the stereo code CS input to the stereo decoding unit 220 is s One-half of (i.e., b s / 2) is used as the bit number b R In the mono decoding unit 210, the mono decoding audio signal x is used M (1),^x M(2),...,^x M (T) is the number of bits decoded M is the number of bits of mono coded CM. Right channel correction coefficient c R It is not the value obtained by equation (1-7-2) itself, but a value greater than 0 and less than 1, when and is the following value: R (1),^y R (2),...,^y R (T) is the number of bits decoded R and for mono decoded sound signal^x M (1),^x M (2),...,^x M (T) is the number of bits decoded M If the same, it is 0.5, the number of bits is b R Bit number b M The more it is, the closer it is to 0 than 0.5. The number of bits b R Bit number b M The smaller it is, the closer it is to 1 than 0.5.

[0149] In addition, the same normalized inner product value candidate and encoding may be used for the left channel and the right channel. Assuming that A and B are the same value, the normalized inner product value candidate r of the left channel stored in the left channel subtraction gain estimation unit 120 and the left channel subtraction gain decoding unit 230 may be the same. Lcand (a) and the encoding Cα corresponding to the candidate cand (a), the candidates r of the normalized inner product value of the right channel stored in the right channel subtraction gain estimation unit 140 and the right channel subtraction gain decoding unit 250 Rcand (b) and the encoding Cβ corresponding to the candidate cand The groups of (b) are the same.

[0150] In addition, the code Cα is actually a code corresponding to the left channel subtraction gain α, and is called the left channel subtraction gain code for the purpose of matching sentences in the description of the encoding device 100 and the decoding device 200, but from the perspective of the code representing the normalized inner product value, it can also be called the left channel inner product code, etc. Similarly, the code Cβ can also be called the right channel inner product code, etc.

[0151] [[Example 2]]

[0152] An example of using a value obtained by also considering the input value of the past frame as the normalized inner product value is described as Example 2. Example 2 does not strictly guarantee the optimality within the frame, that is, the minimization of the energy of the quantization error of the decoded sound signal of the left channel and the minimization of the energy of the quantization error of the decoded sound signal of the right channel, but reduces the rapid change between frames of the left channel subtraction gain α and the rapid change between frames of the right channel subtraction gain β, thereby reducing the noise generated in the decoded sound signal due to the change. That is, in addition to reducing the energy of the quantization error of the decoded sound signal, Example 2 also considers the auditory quality of the decoded sound signal.

[0153] Example 2 is different from Example 1 in the encoding side, that is, the left channel subtraction gain estimation unit 120 and the right channel subtraction gain estimation unit 140, but is the same as Example 1 in the decoding side, that is, the left channel subtraction gain decoding unit 230 and the right channel subtraction gain decoding unit 250. The following description will focus on the differences between Example 2 and Example 1.

[0154] [Left Channel Subtraction Gain Estimation Unit 120]

[0155] like Figure 8 As shown, the left channel subtraction gain estimation unit 120 performs the following steps S120-111 to SS120-113, and steps S120-12 to SS120-14 described in Example 1.

[0156] The left channel subtraction gain estimation unit 120 first uses the input left channel input audio signal x L (1),x L (2),...,x L (T), the input downmix signal x M (1),x M (2),...,x M (T), and the inner product value E used in the previous frame L (-1), the inner product value E used in the current frame is obtained by the following formula (1-8): L (0)(Step S120-111).

[0157] [Formula 14]

[0158]

[0159] Here, ε L is a predetermined value greater than 0 and less than 1, and is stored in advance in the left channel subtraction gain estimation unit 120. In addition, the left channel subtraction gain estimation unit 120 calculates the inner product value E L (0) as the inner product value E used in the previous frame L(-1)" is used in the next frame and stored in the left channel subtraction gain estimation unit 120.

[0160] The left channel subtraction gain estimation unit 120 also uses the input downmix signal x M (1),x M (2),...,x M (T) and the energy E of the downmix signal used in the previous frame M (-1), the energy E of the downmix signal used in the current frame is obtained by the following formula (1-9): M (0)(Step S120-112).

[0161] [Mathematical formula 15]

[0162]

[0163] Here, ε M is a predetermined value greater than 0 and less than 1, and is pre-stored in the left channel subtraction gain estimation unit 120. In addition, the left channel subtraction gain estimation unit 120 calculates the energy E of the downmix signal obtained. M (0) as the energy E of the downmix signal used in the previous frame M (-1)" is used in the next frame and stored in the left channel subtraction gain estimation unit 120.

[0164] Next, the left channel subtraction gain estimation unit 120 uses the inner product value E used in the current frame obtained in step S120-111 L (0) and the energy E of the downmix signal used in the current frame obtained in step S120-112 M (0), the normalized inner product value r is obtained by the following formula (1-10): L (Step S120-113).

[0165] [Formula 16]

[0166] r L =E L (0) / E M (0)…(1-10)

[0167] The left channel subtraction gain estimation unit 120 further performs step S120-12, and then replaces the normalized inner product value r obtained in step S120-11 with L , and use the normalized inner product value r obtained in the above steps S120-113 L To perform step S120-13, and then to perform step S120-14.

[0168] In addition, the above ε Land ε M The closer it is to 1, the smaller the normalized inner product value r L The more likely it is that the input audio signal of the left channel of the past frame and the downmixed signal are included, the normalized inner product value r L , through the normalized inner product value r L The obtained left channel subtraction gain α has a smaller variation between frames.

[0169] [Right Channel Subtraction Gain Estimation Unit 140]

[0170] like Figure 8 As shown, the right channel subtraction gain estimation unit 140 performs the following steps S140-111 to S140-113 and steps S140-12 to S140-14 described in Example 1.

[0171] The right channel subtraction gain estimation unit 140 first uses the input audio signal x of the right channel to obtain the subtraction gain. R (1),x R (2),...,x R (T), the input downmix signal x M (1),x M (2),...,x M (T), and the inner product value E used in the previous frame R (-1), the inner product value E used in the current frame is obtained by the following formula (1-8-2): R (0)(Step S140-111).

[0172] [Mathematical formula 17]

[0173]

[0174] Here, ε R is a predetermined value greater than 0 and less than 1, and is stored in advance in the right channel subtraction gain estimation unit 140. In addition, the right channel subtraction gain estimation unit 140 calculates the inner product value E R (0) is used as the inner product value E used in the previous frame R (-1)" is used in the next frame and stored in the right channel subtraction gain estimation unit 140.

[0175] The right channel subtraction gain estimation unit 140 also uses the input downmix signal x M (1),x M (2),...,x M (T) and the energy E of the downmix signal used in the previous frame M (-1), the energy E of the downmix signal used in the current frame is obtained by formula (1-9) M(0) (Step S140 - 112 ) The right channel subtraction gain estimation unit 140 obtains the energy E of the downmix signal. M (0) as the "energy E of the downmix signal used in the previous frame" M (-1)” and used in the next frame, and stored in the right channel subtraction gain estimation unit 140. In addition, the energy E of the downmix signal used in the current frame is also obtained by the left channel subtraction gain estimation unit 120 through the formula (1-9). M (0), therefore, only one of step S120-112 performed by the left channel subtraction gain estimation unit 120 and step S140-112 performed by the right channel subtraction gain estimation unit 140 may be performed.

[0176] Next, the right channel subtraction gain estimation unit 140 uses the inner product value E used in the current frame obtained in step S140-111 R (0) and the energy E of the downmix signal used in the current frame obtained in step S140-112 M (0), the normalized inner product value r is obtained by the following formula (1-10-2): R (Step S140-113).

[0177] [Mathematical formula 18]

[0178] r R =E R (0) / E M (0)…(1-10-2)

[0179] The right channel subtraction gain estimation unit 140 further performs step S140-12, and then replaces the normalized inner product value r obtained in step S140-11 with R , and use the normalized inner product value r obtained in the above step S140-113 R To perform step S140-13, and then to perform step S140-14.

[0180] In addition, the above ε R and ε M The closer it is to 1, the smaller the normalized inner product value r R The more likely it is that the input audio signal of the right channel of the past frame and the downmixed signal are included, the normalized inner product value r R , through the normalized inner product value r R The obtained right channel subtraction gain β has a smaller variation between frames.

[0181] [Variation of Example 2]

[0182] Regarding Example 2, the same modification as that of the modification of Example 1 with respect to Example 1 can also be performed. This mode will be described as a modification of Example 2. The modification of Example 2 is different from the modification of Example 1 in the encoding side, that is, the left channel subtraction gain estimation unit 120 and the right channel subtraction gain estimation unit 140, but is the same as the modification of Example 1 in the decoding side, that is, the left channel subtraction gain decoding unit 230 and the right channel subtraction gain decoding unit 250. The difference between the modification of Example 1 and the modification of Example 2 is the same as that of Example 2, and therefore, the modification of Example 2 will be described below with reference to the modification of Example 1 and Example 2 as appropriate.

[0183] [Left Channel Subtraction Gain Estimation Unit 120]

[0184] In the left channel subtraction gain estimation unit 120, similarly to the left channel subtraction gain estimation unit 120 of the modification of Example 1, a plurality of sets (A sets, a=1, . . . , A) of candidates for normalized inner product values ​​of the left channel are stored in advance. Lcand (a) and the encoding Cα corresponding to the candidate cand (a) group. Fig. 9 As shown, the left channel subtraction gain estimation unit 120 performs steps S120-111 to S120-113 similar to Example 2, and steps S120-12, S120-15, and S120-16 similar to the modified example of Example 1. Specifically, the steps are as follows.

[0185] The left channel subtraction gain estimation unit 120 first uses the input left channel input audio signal x L (1),x L (2),...,x L (T), the input downmix signal x M (1),x M (2),...,x M (T), and the inner product value E used in the previous frame L (-1), and the inner product value E used in the current frame is obtained by formula (1-8) L (0) (step S120-111). The left channel subtraction gain estimation unit 120 further uses the input downmix signal x M (1),x M (2),...,x M (T) and the energy E of the downmix signal used in the previous frame M (-1), the energy E of the downmix signal used in the current frame is obtained by formula (1-9) M (0) (Step S120-112) Next, the left channel subtraction gain estimation unit 120 uses the inner product value E used in the current frame obtained in step S120-111. L(0) and the energy E of the downmix signal used in the current frame obtained in step S120-112 M (0), and the normalized inner product value r is obtained by formula (1-10) L (Step S120-113) Next, the left channel subtraction gain estimation unit 120 obtains a candidate r of the normalized inner product value with the stored left channel. Lcand (1),...,r Lcand The normalized inner product value r obtained in step S120-113 in (A) L The closest candidate (normalized inner product value r L Quantitative value of L , and get the stored code Cα cand (1),...,Cα cand The closest candidate r in (A) L The corresponding code is used as the left channel subtraction gain code Cα (step S120-15). In addition, the left channel subtraction gain estimation unit 120 uses the left channel difference signal y in the stereo encoding unit 170 L (1),y L (2),...,y L (T) is the number of bits b L , used in the monaural encoding unit 160 to downmix the signal x M (1),x M (2),...,x M (T) is the number of bits b M , and the number of samples per frame T, the left channel correction coefficient c is obtained by formula (1-7) L (Step S120-12) Next, the left channel subtraction gain estimation unit 120 obtains a quantized value of the normalized inner product value obtained in step S120-15. L and the left channel correction coefficient c obtained in step S120-12 L The multiplied value is used as the left channel subtraction gain α (step S120-16).

[0186] [Right Channel Subtraction Gain Estimation Unit 140]

[0187] In the right channel subtraction gain estimation unit 140, similarly to the right channel subtraction gain estimation unit 140 of the modification of Example 1, a plurality of sets (B sets, b=1, . . . , B) of candidates for normalized inner product values ​​of the right channel are stored in advance. Rcand (b) and the encoding Cβ corresponding to the candidate cand (b) group. Fig. 9As shown, the right channel subtraction gain estimation unit 140 performs steps S140-111 to S140-113 similar to Example 2, and steps S140-12, S140-15, and S140-16 similar to the modified example of Example 1. Specifically, the steps are as follows.

[0188] The right channel subtraction gain estimation unit 140 first uses the input audio signal x of the right channel to obtain the subtraction gain. R (1),x R (2),...,x R (T), the input downmix signal x M (1),x M (2),...,x M (T), and the inner product value E used in the previous frame R (-1), through formula (1-8-2), we get the inner product value E used in the current frame R (0) (step S140-111). The right channel subtraction gain estimation unit 140 further uses the input downmix signal x M (1),x M (2),...,x M (T) and the energy E of the downmix signal used in the previous frame M (-1), the energy E of the downmix signal used in the current frame is obtained by formula (1-9) M (0) (Step S140-112) Next, the right channel subtraction gain estimation unit 140 uses the inner product value E used in the current frame obtained in step S140-111. R (0) and the energy E of the downmix signal used in the current frame obtained in step S140-112 M (0), and the normalized inner product value r is obtained by formula (1-10-2) R (Step S140-113) Next, the right channel subtraction gain estimation unit 140 obtains a candidate r of the normalized inner product value with the stored right channel. Rcand (1),...,r Rcand (B) The normalized inner product value r obtained in step S140-113 R The closest candidate (normalized inner product value r R Quantitative value of R , get the stored code Cβ cand (1),...,Cβ cand The closest candidate r in (B) R The corresponding code is used as the right channel subtraction gain code Cβ (step S140-15). In addition, the right channel subtraction gain estimation unit 140 uses the right channel difference signal y in the stereo encoding unit 170R (1),y R (2),...,y R (T) is the number of bits b R , used in the monaural encoding unit 160 to downmix the signal x M (1),x M (2),...,x M (T) is the number of bits b M , and the number of samples per frame T, the right channel correction coefficient c is obtained by formula (1-7-2) R (Step S140-12) Next, the right channel subtraction gain estimation unit 140 obtains a quantized value of the normalized inner product value obtained in step S140-15. R The right channel correction coefficient c obtained in step S140-12 R The multiplied value is used as the right channel subtraction gain β (step S140-16).

[0189] [[Example 3]]

[0190] For example, when the sound, music, etc. included in the input sound signal of the left channel is different from the sound, music, etc. included in the input sound signal of the right channel, since the components of the input sound signal of the left channel can also include the components of the input sound signal of the right channel in the downmix signal, there is a problem that the larger the value used as the left channel subtraction gain α, the more it is heard that the left channel decoded sound signal includes the sound from the input sound signal of the right channel that should not be heard originally, and the larger the value used as the right channel subtraction gain β, the more it is heard that the right channel decoded sound signal includes the sound from the input sound signal of the left channel that should not be heard originally. Therefore, although the energy of the quantization error of the decoded sound signal is not strictly guaranteed to be minimized, the left channel subtraction gain α and the right channel subtraction gain β may be set to values ​​smaller than the values ​​obtained by Example 1 in consideration of auditory quality. In addition, similarly, the left channel subtraction gain α and the right channel subtraction gain β may be set to values ​​smaller than the values ​​obtained by Example 2.

[0191] Specifically, regarding the left channel, in Examples 1 and 2, the normalized inner product value r L and the left channel correction coefficient c L The multiplication value c L × L The quantized value of is set as the left channel subtraction gain α. In Example 3, the normalized inner product value r L , left channel correction coefficient c L and λ which is a predetermined value greater than 0 and less than 1 L The multiplication value λ L ×c L× L The quantized value of is set as the left channel subtraction gain α. Therefore, as in Examples 1 and 2, the multiplication value c L × L As the object of encoding in the left channel subtraction gain estimation unit 120 and decoding in the left channel subtraction gain decoding unit 230, the left channel subtraction gain code Cα represents the multiplication value c L × L The left channel subtraction gain estimation unit 120 and the left channel subtraction gain decoding unit 230 convert the multiplication value c L × L The quantized value of L Multiply and get the left channel subtraction gain α. Alternatively, you can also use the normalized inner product value r L , left channel correction coefficient c L and the preset value λ L The multiplication value λ L ×c L × L The left channel subtraction gain code Cα, which is the object of encoding in the left channel subtraction gain estimation unit 120 and decoding in the left channel subtraction gain decoding unit 230, represents the multiplication value λ. L ×c L × L Quantified value of .

[0192] Similarly, for the right channel, in Examples 1 and 2, the normalized inner product value r R and the right channel correction coefficient c R The multiplication value c R × R The quantized value of is set as the right channel subtraction gain β. In contrast, in Example 3, the normalized inner product value r R , right channel correction coefficient c R and λ which is a predetermined value greater than 0 and less than 1 R The multiplication value λ R ×c R × R The quantized value of is set as the right channel subtraction gain β. Therefore, as in Examples 1 and 2, the multiplication value c R × R As the object of encoding in the right channel subtraction gain estimation unit 140 and decoding in the right channel subtraction gain decoding unit 250, the right channel subtraction gain code Cβ represents the multiplication value c R × R As the quantized value of R × R The quantized value and λ RAlternatively, the normalized inner product value r R , left channel correction coefficient c R and the predetermined value λ R The multiplication value λ R ×c R × R The right channel subtraction gain code Cβ, which is the object of encoding by the right channel subtraction gain estimation unit 140 and decoding by the right channel subtraction gain decoding unit 250, represents the multiplication value λ. R ×c R × R In addition, let λ R is L The same value is sufficient.

[0193] [Variation of Example 3]

[0194] As described above, both the encoding device 100 and the decoding device 200 can calculate the same value. Therefore, similarly to the modification of Example 1 and the modification of Example 2, the normalized inner product value r may be L represents the object of encoding in the left channel subtraction gain estimation unit 120 and decoding in the left channel subtraction gain decoding unit 230, and the left channel subtraction gain code Cα represents the normalized inner product value r L The left channel subtraction gain estimation unit 120 and the left channel subtraction gain decoding unit 230 convert the normalized inner product value r L Quantized value, left channel correction coefficient c L and λ as a predetermined value greater than 0 and less than 1 L Alternatively, the normalized inner product value r L and λ which is a predetermined value greater than 0 and less than 1 L The multiplication value λ L × L As the object of encoding in the left channel subtraction gain estimation unit 120 and decoding in the left channel subtraction gain decoding unit 230, the left channel subtraction gain code Cα represents the multiplication value λ L × L As the quantized value of L × L The quantized value and the left channel correction coefficient c L Multiply them to get the left channel subtraction gain α.

[0195] The same is true for the right channel, the correction coefficient c RThe same value can also be calculated in the encoding device 100 and the decoding device 200. Therefore, the normalized inner product value r may be calculated similarly to the modification of Example 1 and the modification of Example 2. R As the object of encoding by the right channel subtraction gain estimation unit 140 and decoding by the right channel subtraction gain decoding unit 250, the right channel subtraction gain code Cβ represents the normalized inner product value r R As the quantized value of R Quantized value, right channel correction coefficient c R and a predetermined value greater than 0 and less than 1, namely λ R Alternatively, the normalized inner product value r R and a predetermined value greater than 0 and less than 1, namely λ R The multiplication value λ R × R The right channel subtraction gain code Cβ, which is the object of encoding in the right channel subtraction gain estimation unit 140 and decoding in the right channel subtraction gain decoding unit 250, represents the multiplication value λ. R × R The right channel subtraction gain estimation unit 140 and the right channel subtraction gain decoding unit 250 convert the multiplication value λ R × R The quantized value and the right channel correction coefficient c R Multiply them together to obtain the right channel subtraction gain β.

[0196] [[Example 4]]

[0197] The problem of auditory quality described at the beginning of Example 3 occurs when the correlation between the input audio signal of the left channel and the input audio signal of the right channel is small, but this problem hardly occurs when the correlation between the input audio signal of the left channel and the input audio signal of the right channel is large. Therefore, in Example 4, instead of the predetermined value of Example 3, the correlation coefficient of the input audio signal of the left channel and the input audio signal of the right channel, that is, the left-right correlation coefficient γ is used, so that the greater the correlation between the input audio signal of the left channel and the input audio signal of the right channel, the higher the priority is given to reducing the energy of the quantization error of the decoded audio signal, and the smaller the correlation between the input audio signal of the left channel and the input audio signal of the right channel, the higher the priority is given to suppressing the deterioration of auditory quality.

[0198] The encoding side of Example 4 is different from Examples 1 and 2, but the decoding side, that is, the left channel subtraction gain decoding unit 230 and the right channel subtraction gain decoding unit 250 are the same as Examples 1 and 2. The differences between Example 4 and Examples 1 and 2 are described below.

[0199] [Left-right relationship information estimation unit 180]

[0200] like Figure 1 As shown by the middle dotted line, the encoding device 100 of Example 4 further includes a left-right relationship information estimation unit 180. The left-right relationship information estimation unit 180 receives an input audio signal of a left channel and an input audio signal of a right channel of the encoding device 100. The left-right relationship information estimation unit 180 obtains a left-right correlation coefficient γ based on the input audio signal of the left channel and the input audio signal of the right channel, and outputs the result (step S180).

[0201] The left-right correlation coefficient γ is the correlation coefficient between the input audio signal of the left channel and the input audio signal of the right channel, and may also be the sample sequence x of the input audio signal of the left channel. L (1),x L (2),...,x L (T) and the sampling sequence x of the input sound signal of the right channel R (1),x R (2),...,x R The correlation coefficient γ0 of (T) may also be a correlation coefficient that takes into account the time difference, for example, the correlation coefficient γ between the sample sequence of the input audio signal of the left channel and the sample sequence of the input audio signal of the right channel located later than the sample sequence by τ samples. τ .

[0202] Assume that τ is information equivalent to the difference between the arrival time from the sound source that mainly emits sound in a space to the microphone for the left channel and the arrival time from the sound source to the microphone for the right channel (so-called arrival time difference) when the sound signal obtained by AD conversion of the sound collected by the microphone for the left channel arranged in a certain space is the input sound signal for the left channel and the sound signal obtained by AD conversion of the sound collected by the microphone for the right channel arranged in the space is the input sound signal for the right channel, hereinafter referred to as the left-right time difference. The left-right time difference τ can be obtained by any known method, or by the method described in the left-right relationship information estimation unit 181 of the second embodiment. That is, the above-mentioned correlation coefficient γ τ This is information equivalent to the correlation coefficient between the sound signal collected from the sound source to the left channel microphone and the sound signal collected from the sound source to the right channel microphone.

[0203] [Left Channel Subtraction Gain Estimation Unit 120]

[0204] The left channel subtraction gain estimation unit 120 replaces step S120-13 and obtains the normalized inner product value r obtained in step S120-11 or step S120-113. L , the left channel correction coefficient c obtained in step S120-12L , and the left-right correlation coefficient γ obtained in step S180 (step S120-13"). Next, the left channel subtraction gain estimation unit 120 replaces step S120-14 and obtains a candidate α for the stored left channel subtraction gain. cand (1),...,α cand The multiplication value γ×c obtained in step S120-13" in (A) L × L The closest candidate (multiplication value γ×c L × L The quantized value of the left channel is used as the left channel subtraction gain α to obtain the stored code Cα cand (1),...,Cα cand The code corresponding to the left channel subtraction gain α in (A) is used as the left channel subtraction gain Cα (step S120-14").

[0205] [Right Channel Subtraction Gain Estimation Unit 140]

[0206] The right channel subtraction gain estimation unit 140 replaces step S140-13 and obtains the normalized inner product value r obtained in step S140-11 or step S140-113. R , the right channel correction coefficient c obtained in step S140-12 R , and the value obtained by multiplying the left-right correlation coefficient γ obtained in step S180 (step S140-13"). Next, the right channel subtraction gain estimation unit 140 replaces step S140-14 and stores the candidate β of the right channel subtraction gain. cand (1),...,β cand The multiplication value γ×c obtained in step S140-13" in (B) R × R The closest candidate (multiplication value γ×c R × R The quantized value of the right channel is obtained as the right channel subtraction gain β, and the stored code Cβ is obtained. cand (1),...,Cβ cand The code corresponding to the right channel subtraction gain β in (B) is used as the right channel subtraction gain Cβ (step S140-14").

[0207] [[Variation of Example 4]]

[0208] As described above, both the encoding device 100 and the decoding device 200 can calculate the same value. Therefore, the normalized inner product value r L Multiplication value γ×r with left and right correlation coefficient γ LAs the object of encoding in the left channel subtraction gain estimation unit 120 and decoding in the left channel subtraction gain decoding unit 230, the left channel subtraction gain code Cα represents the multiplication value γ×r L The left channel subtraction gain estimation unit 120 and the left channel subtraction gain decoding unit 230 convert the multiplication value γ×r L The quantized value and the left channel correction coefficient c L Multiply them to get the left channel subtraction gain α.

[0209] The same is true for the right channel, the correction coefficient c R The same value can also be calculated in the encoding device 100 and the decoding device 200. Therefore, the normalized inner product value r R Multiplication value γ×r with left and right correlation coefficient γ R As the object of encoding by the right channel subtraction gain estimation unit 140 and decoding by the right channel subtraction gain decoding unit 250, the right channel subtraction gain code Cβ represents the multiplication value γ×r R The right channel subtraction gain estimation unit 140 and the right channel subtraction gain decoding unit 250 convert the multiplication value γ×r R The quantized value and the right channel correction coefficient c R Multiply them together to obtain the right channel subtraction gain β.

[0210] <Second Embodiment>

[0211] The encoding device and decoding device according to the second embodiment will be described.

[0212] 《Encoding device 101》

[0213] like Fig.10 As shown, the encoding device 101 of the second embodiment includes: a downmixing unit 110, a left channel subtraction gain estimation unit 120, a left channel signal subtraction unit 130, a right channel subtraction gain estimation unit 140, a right channel signal subtraction unit 150, a mono encoding unit 160, a stereo encoding unit 170, a left-right relationship information estimation unit 181, and a time shift unit 191. The encoding device 101 of the second embodiment is different from the encoding device 100 of the first embodiment in that the left-right relationship information estimation unit 181 and the time shift unit 191 are included; the left channel subtraction gain estimation unit 120, the left channel signal subtraction unit 130, the right channel subtraction gain estimation unit 140, and the right channel signal subtraction unit 150 use the signal output by the time shift unit 191 instead of the signal output by the downmixing unit 110; and in addition to the above codes, the left-right time difference code Cτ described later is output. The other structures and operations of the encoding device 101 of the second embodiment are the same as those of the encoding device 100 of the first embodiment. The encoding device 101 of the second embodiment performs the following operations on each frame: Fig.11The processing of step S110 to step S191 is illustrated. The following describes the differences between the encoding device 101 of the second embodiment and the encoding device 100 of the first embodiment.

[0214] [Left-right relationship information estimation unit 181]

[0215] The left-right relationship information estimation unit 181 receives an input audio signal of the left channel input to the encoding device 101 and an input audio signal of the right channel input to the encoding device 101. The left-right relationship information estimation unit 181 obtains a left-right time difference τ and a left-right time difference code Cτ which is a code indicating the left-right time difference τ from the input audio signal of the left channel and the input audio signal of the right channel, and outputs them (step S181).

[0216] The left-right time difference τ is information equivalent to the difference between the arrival time from the sound source that mainly emits sound in the space to the microphone for the left channel and the arrival time from the sound source to the microphone for the right channel (so-called arrival time difference), when it is assumed that the sound signal obtained by AD conversion of the sound collected by the microphone for the left channel arranged in a certain space is the input sound signal of the left channel and the sound signal obtained by AD conversion of the sound collected by the microphone for the right channel arranged in the space is the input sound signal of the right channel. In addition, not only the arrival time difference but also information about which microphone arrives earlier is included in the left-right time difference τ, and the left-right time difference τ can take positive and negative values ​​based on either input sound signal. That is, the left-right time difference τ is information indicating in which of the input sound signal of the left channel and the input sound signal of the right channel the same sound signal is included. Hereinafter, when the same sound signal is included in the input sound signal of the left channel before the input sound signal of the right channel, it is also referred to as left channel first, and when the same sound signal is included in the input sound signal of the right channel before the input sound signal of the left channel, it is also referred to as right channel first.

[0217] The left-right time difference τ can also be obtained by any known method. For example, the left-right relationship information estimation unit 181 calculates the predetermined τ max to τ min (For example, τ max is a positive number, τ min is a negative number) cand , calculate the difference between the sample sequence representing the input audio signal of the left channel and the candidate sample number τ that is offset from the sample sequence cand The value of the correlation magnitude of the sample sequence of the input audio signal of the right channel at the position (hereinafter referred to as the correlation value) γ cand , the correlation value γ cand The maximum number of candidate samples τ candThe left-right time difference τ is obtained. That is, in this example, when the left channel is advanced, the left-right time difference τ is a positive value, and when the right channel is advanced, the left-right time difference τ is a negative value. The absolute value of the left-right time difference τ is a value indicating how advanced the leading channel is relative to the other channel (the number of advanced samples). For example, when calculating the correlation value γ using only the samples in the frame cand In the case of τ cand When the value is positive, the partial sampling sequence x of the input audio signal of the right channel is calculated. R (1+τ cand ),x R (2+τ cand ),...,x R (T), and the candidate sampling number τ that deviates forward from the partial sampling sequence cand A partial sample sequence x of the input sound signal of the left channel at the position L (1),x L (2),...,x L (T-τ cand ) is taken as the correlation value γ cand , in τ cand When it is a negative value, the partial sample sequence x of the input sound signal of the left channel is calculated. L (1-τ cand ),x L (2-τ cand ),...,x L (T), and the candidate sampling number -τ that deviates forward from the partial sampling sequence cand A partial sampling sequence x of the input sound signal of the right channel at the position R (1),x R (2),...,x R (T+τ cand ) is taken as the correlation value γ cand Of course, in order to calculate the correlation value γ cand It is also possible to use one or more samples of the past input sound signal that are continuous with the sample sequence of the input sound signal of the current frame. In this case, it is sufficient to store the sample sequence of the input sound signal of the past frame in a predetermined number of frames in a storage unit (not shown) within the left-right relationship information estimation unit 181.

[0218] Alternatively, for example, instead of using the absolute value of the correlation coefficient, the correlation value γ may be calculated using information on the phase of the signal as follows: cand In this example, the left-right relationship information estimation unit 181 first calculates the left channel input audio signal x according to the following equations (3-1) and (3-2). L (1),x L(2),...,x L (T) and the input sound signal x of the right channel R (1),x R (2),...,x R (T) Perform Fourier transform to obtain the spectrum X at each frequency k from 0 to T-1 L (k) and X R (k).

[0219] [Mathematical formula 19]

[0220]

[0221] [Mathematical formula 20]

[0222]

[0223] The left-right relationship information estimation unit 181 uses the obtained spectrum X L (k) and X R (k), the spectrum of the phase difference at each frequency k is obtained by the following formula (3-3):

[0224] [Mathematical formula 21]

[0225]

[0226] By performing inverse Fourier transform on the obtained phase difference spectrum, as shown in the following formula (3-4), for the phase difference spectrum from τ max to τ min The number of candidate samples τ cand , and obtain the phase difference signal ψ(τ cand ).

[0227] [Mathematical formula 22]

[0228]

[0229] The obtained phase difference signal ψ(τ cand ) represents the absolute value of the input sound signal x of the left channel L (1),x L (2),...,x L (T) and the input sound signal x of the right channel R (1),x R (2),...,x R (T) corresponds to a certain correlation, so the candidate sampling number τ cand The phase difference signal ψ(τ cand ) is taken as the correlation value γ candThe left-right relationship information estimation unit 181 obtains the phase difference signal ψ(τ cand ) is the absolute value of the correlation value γ cand Become the maximum candidate sampling number τ cand As the left and right time difference τ. Alternatively, the phase difference signal ψ(τ cand ) is taken as the correlation value γ cand , and using, for example, relative to each τ cand Located relative to the phase difference signal ψ(τ cand ) of the absolute value of τ cand The normalized value is the relative difference between the absolute value of the phase difference signal obtained by each of the plurality of candidate sampling numbers before and after. cand , or you can use a predetermined positive number τ range , the average value is obtained by the following formula (3-5), and the obtained average value ψ is used c (τ cand ) and the phase difference signal ψ(τ cand ), the normalized correlation value obtained by the following formula (3-6) is taken as γ cand use.

[0230] [Mathematical formula 23]

[0231]

[0232] [Mathematical formula 24]

[0233]

[0234] In addition, the normalized correlation value obtained by formula (3-6) is a value greater than 0 and less than 1, which represents τ cand As the time difference between the left and right, the more reasonable it is, the closer it is to 1, τ cand The more unreasonable the left and right time differences are, the closer they are to 0.

[0235] In addition, the left-right relationship information estimation unit 181 encodes the left-right time difference τ using a predetermined encoding method to obtain a code that can uniquely determine the left-right time difference τ, that is, the left-right time difference code Cτ. As the predetermined encoding method, a well-known encoding method such as scalar quantization can be used. In addition, the predetermined number of candidate samples can be either τ or τ. max to τ min The integer values ​​of may also include those in the range from τ max to τ min The fractional value or decimal value between τ max to τ minAny integer value between . In addition, it can be τ max =-τ min , or not. In addition, when a special input audio signal that must be in a certain channel is the object, it can also be τ max and τ min are all positive numbers, or τ max and τ min All are negative numbers.

[0236] In addition, when the encoding device 101 estimates the subtraction gain based on the principle of minimizing the quantization error of Example 4 or the modified example of Example 4 described in the first embodiment, the left-right relationship information estimation unit 181 also outputs the correlation value between the sample sequence of the input audio signal of the left channel and the sample sequence of the input audio signal of the right channel at a position that is offset from the sample sequence by the left-right time difference τ, that is, the correlation value between the sample sequence of the input audio signal of the left channel and the sample sequence of the input audio signal of the right channel at a position that is offset from the sample sequence by the left-right time difference τ. max to τ min The number of candidate samples τ cand The calculated correlation value γ cand The maximum value among them is taken as the left-right correlation coefficient γ (step S180).

[0237] [Time Shift Section 191]

[0238] The downmixed signal x output from the downmixing unit 110 is input to the time shift unit 191. M (1),x M (2),...,x M (T) and the left-right time difference τ output by the left-right relationship information estimation unit 181. When the left-right time difference τ is a positive value (that is, the left-right time difference τ indicates that the left channel is ahead), the time shift unit 191 shifts the downmix signal x M (1),x M (2),...,x M (T) is directly output to the left channel subtraction gain estimation unit 120 and the left channel signal subtraction unit 130 (i.e., it is determined to be used in the left channel subtraction gain estimation unit 120 and the left channel signal subtraction unit 130), and the downmix signal is delayed by |τ| samples (the number of samples of the absolute value of the left and right time difference τ, the number of samples of the magnitude represented by the left and right time difference τ) to obtain a signal x M (1-|τ|),x M (2-|τ|),...,x M (T-|τ|) is the delayed downmix signal x M' (1),x M' (2),...,x M'(T) is output to the right channel subtraction gain estimation unit 140 and the right channel signal subtraction unit 150 (i.e., is determined to be used in the right channel subtraction gain estimation unit 140 and the right channel signal subtraction unit 150), and when the left-right time difference τ is a negative value (i.e., the left-right time difference τ indicates that the right channel is ahead), the downmix signal is delayed by |τ| samples to obtain a signal x M (1-|τ|),x M (2-|τ|),...,x M (T-|τ|) is the delayed downmix signal x M' (1),x M' (2),...,x M' (T) is output to the left channel subtraction unit 120 and the left channel signal subtraction unit 130 (that is, it is determined to be used in the left channel subtraction unit 120 and the left channel signal subtraction unit 130), and the downmix signal x M (1),x M (2),...,x M (T) is directly output to the right channel subtraction gain estimation unit 140 and the right channel signal subtraction unit 150 (i.e., it is determined to be used in the right channel subtraction gain estimation unit 140 and the right channel signal subtraction unit 150), and when the left-right time difference τ is 0 (i.e., the left-right time difference τ indicates that neither channel is advanced), the downmix signal x M (1),x M (2),...,x M (T) is directly output to the left channel subtraction gain estimation unit 120, the left channel signal subtraction unit 130, the right channel subtraction gain estimation unit 140 and the right channel signal subtraction unit 150 (i.e., it is determined to be used in the left channel subtraction gain estimation unit 120, the left channel signal subtraction unit 130, the right channel subtraction gain estimation unit 140 and the right channel signal subtraction unit 150) (step S191). That is, for the channel with a shorter arrival time among the left channel and the right channel, the input downmix signal is directly output to the subtraction gain estimation unit and the signal subtraction unit of the channel, and for the channel with a longer arrival time among the left channel and the right channel, the signal obtained by delaying the input downmix signal by the absolute value |τ| of the left and right time difference τ is output to the subtraction gain estimation unit and the signal subtraction unit of the channel. In order to obtain the delayed downmix signal in the time shift unit 191 , the downmix signal of the past frame is used. Therefore, a storage unit (not shown) in the time shift unit 191 stores the downmix signal input in the past frame for a predetermined number of frames.

[0239] [Left channel subtraction gain estimation unit 120, left channel signal subtraction unit 130, right channel subtraction gain estimation unit 140, right channel signal subtraction unit 150]

[0240] The left channel subtraction gain estimation section 120, the left channel signal subtraction section 130, the right channel subtraction gain estimation section 140, and the right channel signal subtraction section 150 use the downmix signal x input from the time shift section 191. M (1),x M (2),...,x M (T) or delayed downmix signal x M' (1),x M' (2),...,x M' (T) The same operation as described in the first embodiment is performed, but the downmix signal x output by the downmixing unit 110 is replaced with M (1),x M (2),...,x M That is, the left channel subtraction gain estimation unit 120, the left channel signal subtraction unit 130, the right channel subtraction gain estimation unit 140 and the right channel signal subtraction unit 150 use the downmix signal x determined by the time shift unit 191. M (1),x M (2),...,x M (T) or delayed downmix signal x M' (1),x M' (2),...,x M' (T) The same operation as described in the first embodiment is performed.

[0241] 《Decoding Device 201》

[0242] like Fig.12 As shown, the decoding device 201 of the second embodiment includes: a mono decoding unit 210, a stereo decoding unit 220, a left channel subtraction gain decoding unit 230, a left channel signal adding unit 240, a right channel signal subtraction gain decoding unit 250, a right channel signal adding unit 260, a left-right time difference decoding unit 271, and a time shift unit 281. The difference between the decoding device 201 of the second embodiment and the decoding device 200 of the first embodiment is that in addition to the above-mentioned codes, the left-right time difference code Cτ described later is also input; the left-right time difference decoding unit 271 and the time shift unit 281 are included; and the left channel signal adding unit 240 and the right channel signal adding unit 260 use the signal output by the time shift unit 281 to replace the signal output by the mono decoding unit 210. The other structures and operations of the decoding device 201 of the second embodiment are the same as those of the decoding device 200 of the first embodiment. The decoding device 201 of the second embodiment performs the following operations on each frame: Fig.13 The processing from step S210 to step S281 is illustrated. The following describes the differences between the decoding apparatus 201 of the second embodiment and the decoding apparatus 200 of the first embodiment.

[0243] [Left-right time difference decoding unit 271]

[0244] The left-right time difference code Cτ input to the decoding device 201 is input to the left-right time difference decoding unit 271. The left-right time difference decoding unit 271 decodes the left-right time difference code Cτ in a predetermined decoding method, obtains the left-right time difference τ and outputs it (step S271). As the predetermined decoding method, a decoding method corresponding to the encoding method used in the left-right relationship information estimation unit 181 of the corresponding encoding device 101 is used. The left-right time difference τ obtained by the left-right time difference decoding unit 271 is the same value as the left-right time difference τ obtained by the left-right relationship information estimation unit 181 of the corresponding encoding device 101, which is obtained from τ max to τ min Any value in the range of .

[0245] [Time Shift Section 281]

[0246] The monaural decoded audio signal ^x output by the monaural decoding unit 210 is input to the time shift unit 281. M (1),^x M (2),...,^x M (T) and the left-right time difference τ output by the left-right time difference decoding unit 271. When the left-right time difference τ is a positive value (that is, the left-right time difference τ indicates that the left channel is ahead), the time shift unit 281 converts the monaural decoded audio signal ^x M (1),^x M (2),...,^x M (T) is directly output to the left channel signal adding section 240 (ie, is determined to be used in the left channel signal adding section 240), and the signal ^x obtained by delaying the monaural decoded audio signal by |τ| samples M (1-|τ|),^x M (2-|τ|),...,^x M (T-|τ|) is the delayed mono decoded sound signal^x M' (1),^x M' (2),...,^x M' (T) is output to the right channel signal adding unit 260 (i.e., it is determined to be used in the right channel signal adding unit 260), and when the left-right time difference τ is a negative value (i.e., the left-right time difference τ indicates that the right channel is ahead), the signal ^x obtained by delaying the monaural decoded audio signal by |τ| samples is M (1-|τ|),^x M (2-|τ|),...,^x M (T-|τ|) is the delayed mono decoded sound signal^x M' (1),^x M'(2),...,^x M' (T) is output to the left channel signal adding section 240 (ie, is determined to be used in the left channel signal adding section 240), and the monaural decoded audio signal is converted into M (1),^x M (2),...,^x M (T) is directly output to the right channel signal adding unit 260 (i.e., it is determined to be used in the right channel signal adding unit 260), and when the left and right time difference τ is 0 (i.e., the left and right time difference τ indicates that neither channel is advanced), the mono decoded audio signal ^x M (1),^x M (2),...,^x M (T) is directly output to the left channel signal adding unit 240 and the right channel signal adding unit 260 (i.e., it is determined to be used in the left channel signal adding unit 240 and the right channel signal adding unit 260) (step S281). In addition, since the monaural decoded sound signal of the past frame is used in the time shift unit 281 to obtain the delayed monaural decoded sound signal, the monaural decoded sound signal input in the past frame is stored in a storage unit (not shown) in the time shift unit 281 for a predetermined number of frames.

[0247] [Left channel signal adding unit 240, right channel signal adding unit 260]

[0248] The left channel signal adding section 240 and the right channel signal adding section 260 use the monaural decoded audio signal input from the time shift section 281. M (1),^x M (2),...,^x M (T) or delayed mono decoded sound signal^x M' (1),^x M' (2),...,^x M' (T) The same operation as described in the first embodiment is performed, but the monaural decoded audio signal ^x outputted by the monaural decoding unit 210 is replaced by M (1),^x M (2),...,^x M (T) (steps S240, S260). That is, the left channel signal adding unit 240 and the right channel signal adding unit 260 use the monaural decoded audio signal ^x determined by the time shift unit 281. M (1),^x M (2),...,^x M (T) or delayed mono decoded sound signal^x M' (1),^x M' (2),...,^x M'(T) The same operation as described in the first embodiment is performed.

[0249] <Third Embodiment>

[0250] The encoding device 101 of the second embodiment may be modified to generate a downmix signal by taking into account the relationship between the input audio signal of the left channel and the input audio signal of the right channel, and this method is described as a third embodiment. In addition, the encoding obtained by the encoding device of the third embodiment can be decoded by the decoding device 201 of the second embodiment, so the description of the decoding device is omitted.

[0251] 《Encoding device 102》

[0252] like Fig.10 As shown, the encoding device 102 of the third embodiment includes: a downmixing unit 112, a left channel subtraction gain estimation unit 120, a left channel signal subtraction unit 130, a right channel subtraction gain estimation unit 140, a right channel signal subtraction unit 150, a mono encoding unit 160, a stereo encoding unit 170, a left-right relationship information estimation unit 182, and a time shift unit 191. The encoding device 102 of the third embodiment is different from the encoding device 101 of the second embodiment in that the left-right relationship information estimation unit 182 is included instead of the left-right relationship information estimation unit 181, and the downmixing unit 112 is included instead of the downmixing unit 110. Fig.10 As shown by the middle dotted line, the left-right relationship information estimation unit 182 obtains and outputs the left-right correlation coefficient γ and the preceding channel information, and the output left-right correlation coefficient γ and the preceding channel information are input to the downmixing unit 112 for use. The other structures and operations of the encoding device 102 of the third embodiment are the same as those of the encoding device 101 of the second embodiment. The encoding device 102 of the third embodiment performs the following steps on each frame: Fig.14 The processing of step S112 to step S191 is illustrated. The following describes the differences between the encoding device 102 of the third embodiment and the encoding device 101 of the second embodiment.

[0253] [Left-right relationship information estimation unit 182]

[0254] The left-right relationship information estimation unit 182 receives the input audio signal of the left channel input to the encoding device 102 and the input audio signal of the right channel input to the encoding device 102. The left-right relationship information estimation unit 182 obtains the left-right time difference τ, the left-right time difference code Cτ as a code representing the left-right time difference τ, the left-right correlation coefficient γ, and the preceding channel information based on the input audio signal of the left channel and the input audio signal of the right channel, and outputs them (step S182). The processing of obtaining the left-right time difference τ and the left-right time difference code Cτ by the left-right relationship information estimation unit 182 is the same as that of the left-right relationship information estimation unit 181 of the second embodiment.

[0255] The left-right correlation coefficient γ is information corresponding to the correlation coefficient between the sound signal collected by the microphone for the left channel from the sound source and the sound signal collected by the microphone for the right channel from the sound source, as assumed in the above description of the left-right relationship information estimation unit 181 of the second embodiment. The preceding channel information is information corresponding to which microphone the sound emitted by the sound source reaches first, and is information indicating whether the same sound signal is included first in the input sound signal of the left channel and the input sound signal of the right channel, and is information indicating which of the left channel and the right channel precedes.

[0256] According to the example described above in the description of the left-right relationship information estimation unit 181 of the second embodiment, the left-right relationship information estimation unit 182 calculates the correlation value between the sample sequence of the input sound signal of the left channel and the sample sequence of the input sound signal of the right channel at a position later than the sample sequence by the left-right time difference τ, that is, the correlation value between the sample sequence of the input sound signal of the left channel and the sample sequence of the input sound signal of the right channel at a position later than the sample sequence by the left-right time difference τ. max to τ min The number of candidate samples τ cand The calculated correlation value γ cand The maximum value of is output as the left-right correlation coefficient γ. In addition, when the left-right time difference τ is a positive value, the left-right relationship information estimation unit 182 obtains and outputs information indicating that the left channel precedes as the preceding channel information, and when the left-right time difference τ is a negative value, the left-right relationship information estimation unit 182 obtains and outputs information indicating that the right channel precedes as the preceding channel information. When the left-right time difference τ is 0, the left-right relationship information estimation unit 182 may obtain and output information indicating that the left channel precedes as the preceding channel information, may obtain and output information indicating that the right channel precedes as the preceding channel information, or may obtain and output information indicating that neither channel precedes as the preceding channel information.

[0257] [Downmixing unit 112]

[0258] The downmixing unit 112 is input with the input audio signal of the left channel input to the encoding device 102, the input audio signal of the right channel input to the encoding device 102, the left-right correlation coefficient γ output by the left-right correlation information estimation unit 182, and the preceding channel information output by the left-right correlation information estimation unit 182. The downmixing unit 112 obtains a downmixed signal by weightedly averaging the input audio signal of the left channel and the input audio signal of the right channel, and outputs the downmixed signal so that the larger the left-right correlation coefficient γ is, the larger the input audio signal of the preceding channel of the input audio signal of the left channel and the input audio signal of the right channel is included in the downmixed signal (step S112).

[0259] For example, if the absolute value or normalized value of the correlation coefficient is used as the correlation value as in the above example in the description of the left-right relationship information estimation unit 181 of the second embodiment, the obtained left-right correlation coefficient γ is a value not less than 0 and not more than 1. Therefore, the downmixing unit 112 uses the weight determined by the left-right correlation coefficient γ for each corresponding sample number t to mix the input audio signal x of the left channel. L (t) and the input sound signal x of the right channel R (t) The value obtained by weighted addition is taken as the downmix signal x M Specifically, when the preceding channel information is information indicating that the left channel precedes, that is, when the left channel precedes, the downmixing unit 112 sets x M (t) = ((1 + γ) / 2) × x L (t)+((1-γ) / 2)×x R (t), when the preceding channel information is information indicating that the right channel precedes, that is, when the right channel precedes, let x M (t) = ((1-γ) / 2) × x L (t)+((1+γ) / 2)×x R (t), and get the downmixed signal x M The downmixed signal is obtained by the downmixing unit 112 in this way, so that the smaller the left-right correlation coefficient γ is, that is, the smaller the correlation between the input sound signal of the left channel and the input sound signal of the right channel is, the closer the downmixed signal is to a signal obtained by averaging the input sound signal of the left channel and the input sound signal of the right channel, and the larger the left-right correlation coefficient γ is, that is, the larger the correlation between the input sound signal of the left channel and the input sound signal of the right channel is, the closer the downmixed signal is to the input sound signal of the preceding channel of the input sound signal of the left channel and the input sound signal of the right channel.

[0260] In addition, when none of the channels is advanced, the downmixing unit 112 may average the input audio signal of the left channel and the input audio signal of the right channel so that the input audio signal of the left channel and the input audio signal of the right channel are included in the downmixed signal with the same weight, and output the downmixed signal. Therefore, when the advance channel information indicates that none of the channels is advanced, the downmixing unit 112 averages the input audio signal of the left channel and the input audio signal of the right channel for each sample number t. L (t) and the input sound signal x of the right channel R (t) The averaged x M (t) = (x L (t)+x R (t)) / 2 is set as the downmix signal x M (t).

[0261] <Fourth Embodiment>

[0262] The encoding device 100 of the first embodiment may be modified to generate a downmix signal by taking into account the relationship between the input audio signal of the left channel and the input audio signal of the right channel, and this method will be described as a fourth embodiment. In addition, since the code obtained by the encoding device of the fourth embodiment can be decoded by the decoding device 200 of the first embodiment, the description of the decoding device is omitted.

[0263] 《Encoding device 103》

[0264] like Figure 1 As shown in FIG. 1 , the encoding device 103 of the fourth embodiment includes: a downmixing unit 112, a left channel subtraction gain estimation unit 120, a left channel signal subtraction unit 130, a right channel subtraction gain estimation unit 140, a right channel signal subtraction unit 150, a mono encoding unit 160, a stereo encoding unit 170, and a left-right relationship information estimation unit 183. The encoding device 103 of the fourth embodiment is different from the encoding device 100 of the first embodiment in that the downmixing unit 110 is replaced by the downmixing unit 112. Figure 1 The dotted line in the middle includes a left-right relationship information estimation unit 183, which obtains and outputs the left-right correlation coefficient γ and the preceding channel information. The output left-right correlation coefficient γ and the preceding channel information are input to the downmixing unit 112 for use. The other structures and operations of the encoding device 103 of the fourth embodiment are the same as those of the encoding device 100 of the first embodiment. In addition, the operation of the downmixing unit 112 of the encoding device 103 of the fourth embodiment is the same as the operation of the downmixing unit 112 of the encoding device 102 of the third embodiment. The encoding device 103 of the fourth embodiment performs the following operations on each frame: Fig.15 The processing of step S112 to step S183 is illustrated. Hereinafter, the points in which the encoding device 103 according to the fourth embodiment is different from the encoding device 100 according to the first embodiment and the encoding device 102 according to the third embodiment will be described.

[0265] [Left-right relationship information estimation unit 183]

[0266] The left-channel input audio signal input to the encoding device 103 and the right-channel input audio signal input to the encoding device 103 are input to the left-channel relationship information estimation unit 183. The left-channel relationship information estimation unit 183 obtains a left-right correlation coefficient γ and preceding channel information based on the input audio signal of the left channel and the input audio signal of the right channel, and outputs them (step S183).

[0267] The left-right correlation coefficient γ and the preceding channel information obtained and output by the left-right relationship information estimation unit 183 are the same as those described in the third embodiment. That is, except that the left-right relationship information estimation unit 183 may not obtain the left-right time difference τ and the left-right time difference code Cτ and does not output them, they are the same as the left-right relationship information estimation unit 182.

[0268] For example, regarding the max to τ min The number of candidate samples τ cand The left-right relationship information estimation unit 183 compares the sample sequence of the input audio signal of the left channel with the sample sequence shifted backward by each candidate sample number τ. cand The correlation value γ of the sampling sequence of the input sound signal of the right channel at the position cand The maximum value in is taken as the left and right correlation coefficient γ and output. When the correlation value is the maximum value, τ cand When the correlation value is positive, the information indicating the left channel is obtained as the preceding channel information and output. cand When the correlation value is a negative value, the information indicating the right channel precedence is obtained and output as the precedent channel information. cand When it is 0, the left-right relationship information estimation unit 183 may obtain and output information indicating that the left channel precedes as the preceding channel information, may obtain and output information indicating that the right channel precedes as the preceding channel information, or may obtain and output information indicating that neither channel precedes as the preceding channel information.

[0269] <Program and Recording Medium>

[0270] The processing of each part of each encoding device and each decoding device can also be realized by a computer. In this case, the processing content of the function that each device should have is recorded in a program. Then, Fig.16 The storage unit 1020 of the computer shown reads the program and operates the arithmetic processing unit 1010, the input unit 1030, the output unit 1040, etc., thereby realizing various processing functions in the above-mentioned devices on the computer.

[0271] The program describing the processing contents can be recorded in a computer-readable recording medium. The computer-readable recording medium is, for example, a non-transitory recording medium, specifically, a magnetic recording device, an optical disk, and the like.

[0272] The program may be circulated by selling, transferring, renting, or other removable recording media such as DVDs and CD-ROMs that record the program. Furthermore, the program may be stored in a storage device of a server computer and circulated by forwarding the program from the server computer to other computers via a network.

[0273] The computer executing such a program, for example, first temporarily stores the program recorded in the transportable recording medium or the program forwarded from the server computer in its own non-temporary storage device, namely the auxiliary recording unit 1050. Then, when executing the process, the storage unit 1020 reads the program stored in its own non-temporary storage device, namely the auxiliary recording unit 1050, and executes the process according to the read program. Moreover, as another execution method of the program, the computer can also directly read the program from the transportable recording medium into the storage unit 1020 and execute the process according to the program, and further, it can also execute the process according to the received program one by one each time the program is forwarded from the server computer to the computer. Moreover, it can also be set as a structure that performs the above-mentioned process by not forwarding the program from the server computer to the computer, and only realizing the processing function through the execution instruction and the result acquisition, so-called ASP (Application Service Provider) type service. Moreover, the program in this method contains information for the processing of the electronic computer, that is, information based on the program (although not a direct instruction to the computer, but having the nature of specifying the processing of the computer, etc.).

[0274] Furthermore, in this embodiment, the present device is configured by executing a predetermined program on a computer, but at least a part of the processing contents may be realized by hardware.

[0275] Furthermore, it is of course possible to make appropriate changes within the scope not departing from the gist of the present invention.

Claims

1. A method for encoding a sound signal, which encodes an input sound signal frame by frame, characterized in that: include: A downmixing step, obtaining a signal obtained by mixing the input left channel input sound signal and the input right channel input sound signal, that is, a downmixed signal; A left channel subtraction gain estimation step, obtaining a left channel subtraction gain α and a code representing the left channel subtraction gain α, namely, a left channel subtraction gain code Cα, according to the left channel input sound signal and the downmix signal; The left channel signal subtraction step obtains the sample value x of the sound signal input from the left channel for each corresponding sample t. L (t) by subtracting the sample value x of the downmix signal M (t) The value x obtained by multiplying the value of the left channel subtraction gain α L (t)-α×x M The sequence of (t) is taken as the left channel difference signal; a right channel subtraction gain estimation step, obtaining a right channel subtraction gain β and a code representing the right channel subtraction gain β, i.e., a right channel subtraction gain code Cβ, according to the right channel input sound signal and the downmix signal; The right channel signal subtraction step obtains the sampling value x of the sound signal input from the right channel for each corresponding sample t. R (t) by subtracting the sample value x of the downmix signal M The value x obtained by multiplying (t) and the right channel subtraction gain β R (t)-β×x M The sequence of (t) is used as the right channel difference signal; a mono encoding step of encoding the downmix signal to obtain a mono code CM; and A stereo encoding step is to encode the left channel differential signal and the right channel differential signal to obtain a stereo code CS, The number of bits used for encoding the downmix signal in the mono encoding step is set to b M , the number of bits used for encoding the left channel differential signal in the stereo encoding step is set to b L , the number of bits used for encoding the right channel differential signal in the stereo encoding step is set to b R , In the left channel subtraction gain estimation step, Get the left channel correction coefficient c L and the normalized inner product value r of the downmix signal for the left channel input sound signal L The quantized value of the multiplication value is used as the left channel subtraction gain α, and the inner product value r with the left channel subtraction gain α or the normalized inner product value r is obtained. L The quantized value corresponding to the encoding is used as the left channel subtraction gain encoding Cα, the left channel correction coefficient c L is a value greater than 0 and less than 1, and is in b L =b M When 0.5, b L B M The more it is, the closer it is to 0 than 0.

5. L B M The smaller the value, the closer it is to 1 than 0.

5. In the right channel subtraction gain estimation step, Get the right channel correction coefficient c R and the normalized inner product value r of the downmix signal for the right channel input sound signal R The quantized value of the multiplication value is used as the right channel subtraction gain β, and the inner product value r with the right channel subtraction gain β or the normalized inner product value r is obtained. R The quantized value corresponding to the encoding is used as the right channel subtraction gain encoding Cβ, the right channel correction coefficient c R is a value greater than 0 and less than 1, and is in b R =b M When 0.5, b R B M The more it is, the closer it is to 0 than 0.

5. R B M The smaller the value, the closer it is to 1 than 0.

5.

2. A sound signal encoding method, which encodes an input sound signal frame by frame, characterized in that: include: A downmixing step, obtaining a signal obtained by mixing the input left channel input sound signal and the input right channel input sound signal, that is, a downmixed signal; A left channel subtraction gain estimation step, obtaining a left channel subtraction gain α and a code representing the left channel subtraction gain α, namely, a left channel subtraction gain code Cα, according to the left channel input sound signal and the downmix signal; The left channel signal subtraction step obtains the sample value x of the sound signal input from the left channel for each corresponding sample t. L (t) by subtracting the sample value x of the downmix signal M (t) The value x obtained by multiplying the value of the left channel subtraction gain α L (t)-α×x M The sequence of (t) is taken as the left channel difference signal; a right channel subtraction gain estimation step, obtaining a right channel subtraction gain β and a code representing the right channel subtraction gain β, i.e., a right channel subtraction gain code Cβ, according to the right channel input sound signal and the downmix signal; The right channel signal subtraction step obtains the sampling value x of the sound signal input from the right channel for each corresponding sample t. R (t) by subtracting the sample value x of the downmix signal M The value x obtained by multiplying (t) and the right channel subtraction gain β R (t)-β×x M The sequence of (t) is used as the right channel difference signal; a mono encoding step of encoding the downmix signal to obtain a mono code CM; and A stereo encoding step is to encode the left channel differential signal and the right channel differential signal to obtain a stereo code CS, The number of bits used for encoding the downmix signal in the mono encoding step is set to b M , the number of bits used for encoding the left channel differential signal in the stereo encoding step is set to b L , the number of bits used for encoding the right channel differential signal in the stereo encoding step is set to b R , In the left channel subtraction gain estimation step, Get the left channel correction coefficient c L , a normalized inner product value r of the downmix signal for the left channel input sound signal L , and a pre-set value greater than 0 and less than 1, that is, a quantized value of the multiplication value of the left channel coefficient value as the left channel subtraction gain α, and obtain the inner product value r with the left channel subtraction gain α and the normalized inner product value r L The quantized value, or the normalized inner product value r L The code corresponding to the quantized value of the value multiplied by the left channel coefficient value is used as the left channel subtraction gain code Cα, and the left channel correction coefficient c L is a value greater than 0 and less than 1, and is in b L =b M When 0.5, b L B M The more it is, the closer it is to 0 than 0.

5. L B M The smaller the value, the closer it is to 1 than 0.

5. In the right channel subtraction gain estimation step, Get the right channel correction coefficient c R , a normalized inner product value r of the downmix signal for the right channel input sound signal R , and a preset value greater than 0 and less than 1, that is, a quantized value of the multiplication value of the right channel coefficient value as the right channel subtraction gain β, and obtain the inner product value r with the right channel subtraction gain β and the normalized inner product value r R The quantized value or the normalized inner product value r R The code corresponding to the quantized value of the value multiplied by the coefficient value of the right channel is used as the right channel subtraction gain code Cβ, and the right channel correction coefficient c R is a value greater than 0 and less than 1, and is in b R =b M When 0.5, b R B M The more it is, the closer it is to 0 than 0.

5. R B M The smaller the value, the closer it is to 1 than 0.

5.

3. A sound signal encoding method, which encodes an input sound signal frame by frame, characterized in that: include: A downmixing step, obtaining a signal obtained by mixing the input left channel input sound signal and the input right channel input sound signal, that is, a downmixed signal; A left channel subtraction gain estimation step, obtaining a left channel subtraction gain α and a code representing the left channel subtraction gain α, namely, a left channel subtraction gain code Cα, according to the left channel input sound signal and the downmix signal; The left channel signal subtraction step obtains the sample value x of the sound signal input from the left channel for each corresponding sample t. L (t) by subtracting the sample value x of the downmix signal M (t) The value x obtained by multiplying the value of the left channel subtraction gain α L (t)-α×x M The sequence of (t) is taken as the left channel difference signal; a right channel subtraction gain estimation step, obtaining a right channel subtraction gain β and a code representing the right channel subtraction gain β, i.e., a right channel subtraction gain code Cβ, according to the right channel input sound signal and the downmix signal; The right channel signal subtraction step obtains the sampling value x of the sound signal input from the right channel for each corresponding sample t. R (t) by subtracting the sample value x of the downmix signal M The value x obtained by multiplying (t) and the right channel subtraction gain β R (t)-β×x M The sequence of (t) is used as the right channel difference signal; a mono encoding step of encoding the downmix signal to obtain a mono code CM; and A stereo encoding step is to encode the left channel differential signal and the right channel differential signal to obtain a stereo code CS, The number of bits used for encoding the downmix signal in the mono encoding step is set to b M , the number of bits used for encoding the left channel differential signal in the stereo encoding step is set to b L , the number of bits used for encoding the right channel differential signal in the stereo encoding step is set to b R , In the left channel subtraction gain estimation step, Get the left channel correction coefficient c L , a normalized inner product value r of the downmix signal for the left channel input sound signal L , the quantized value of the multiplication value of the left channel coefficient value determined for each frame, which is greater than or equal to 0 and less than or equal to 1, is used as the left channel subtraction gain α, and the normalized inner product value r is obtained. L The quantized value, or the normalized inner product value r L The code corresponding to the quantized value of the value multiplied by the left channel coefficient value is used as the left channel subtraction gain code Cα, and the left channel correction coefficient c L is a value greater than 0 and less than 1, and is in b L =b M When 0.5, b L B M The more it is, the closer it is to 0 than 0.

5. L B M The smaller the value, the closer it is to 1 than 0.

5. In the right channel subtraction gain estimation step, Get the right channel correction coefficient c R , a normalized inner product value r of the downmix signal for the right channel input sound signal R , the quantized value of the multiplication value of the coefficient value for the right channel determined for each frame, which is greater than or equal to 0 and less than or equal to 1, is used as the right channel subtraction gain β, and the normalized inner product value r is obtained with the right channel subtraction gain β and the normalized inner product value r R The quantized value, or the normalized inner product value r R The code corresponding to the quantized value of the value multiplied by the coefficient value of the right channel is used as the right channel subtraction gain code Cβ, and the right channel correction coefficient c R is a value greater than 0 and less than 1, and is in b R =b M When 0.5, b R B M The more it is, the closer it is to 0 than 0.

5. R B M The smaller the value, the closer it is to 1 than 0.

5.

4. The sound signal encoding method according to claim 3, characterized in that: The method further includes a left-right correlation estimation step, in which a left-right correlation coefficient is obtained as a correlation coefficient between the left channel input sound signal and the right channel input sound signal. In the left channel subtraction gain estimation step, the left-right correlation coefficient is used as the left channel coefficient value, In the right channel subtraction gain estimation step, the left-right correlation coefficient is used as the coefficient value for the right channel.

5. The sound signal encoding method according to any one of claims 1 to 3, characterized in that: The method further includes a left-right relationship information estimation step, in which information indicating which channel of the left channel and the right channel precedes, namely, preceding channel information, and a correlation coefficient between the left channel input sound signal and the right channel input sound signal, namely, a left-right correlation coefficient, are obtained. In the downmixing step, Based on the preceding channel information and the left-right correlation coefficient, the left channel input sound signal and the right channel input sound signal are weighted averaged to obtain the downmixed signal in such a manner that the larger the left-right correlation coefficient, the greater the inclusion of the preceding channel input sound signal in the left channel input sound signal and the right channel input sound signal.

6. The sound signal encoding method according to any one of claims 1 to 3, characterized in that: Set the number of samples per frame to T, The left channel correction coefficient c L for [Mathematical formula 25] The right channel correction coefficient c R for [Mathematical formula 26] 7. The sound signal encoding method according to any one of claims 1 to 3, characterized in that: ε L , ε R , ε M are values ​​greater than 0 and less than 1 respectively. In the left channel subtraction gain estimation step, The inner product value E of the left channel input sound signal, the downmixed signal and the previous frame is used. L (-1), through [Mathematical formula 27] Get the inner product value E L (0), Use the energy E of the downmix signal and the downmix signal of the previous frame M (-1), through [Mathematical formula 28] Get the energy E of the downmix signal M (0), The inner product value E will be used L (0) and the energy E of the downmix signal M (0), through [Mathematical formula 29] r L =And L (0) / E M (0) The obtained r L The inner product value normalized for the left channel input sound signal as the downmix signal, In the right channel subtraction gain estimation step, The inner product value E of the right channel input sound signal, the downmixed signal and the previous frame is used. R (-1), through [Math 30] Get the inner product value E L (0), Use the energy E of the downmix signal and the downmix signal of the previous frame M (-1), By [Mathematical formula 31] Get the energy E of the downmix signal M (0), The inner product value E will be used L (0) and the energy E of the downmix signal M (0), through [Mathematical formula 32] r R =And R (0) / E M (0) The obtained r R The inner product value normalized with respect to the right channel input sound signal is used as the downmix signal.

8. The sound signal encoding method according to any one of claims 1 to 3, characterized in that: Also includes: a left-right time difference estimation step, obtaining a left-right time difference τ and a code representing the left-right time difference τ, i.e., a left-right time difference code Cτ, according to the left channel input sound signal and the right channel input sound signal; and a time shifting step, in which, when the left-right time difference τ indicates that the left channel is ahead, the downmixed signal is directly used in the left channel subtraction gain estimation step and the left channel signal subtraction step, and a delayed downmixed signal is used in the right channel subtraction gain estimation step and the right channel signal subtraction step, wherein the delayed downmixed signal is a signal in which the downmixed signal is delayed by a magnitude corresponding to the left-right time difference τ, When the left-right time difference τ indicates that the right channel is ahead, it is determined to directly use the downmix signal in the right channel subtraction gain estimation step and the right channel signal subtraction step, and it is determined to use a delayed downmix signal in the left channel subtraction gain estimation step and the left channel signal subtraction step, the delayed downmix signal being a signal in which the downmix signal is delayed by a magnitude corresponding to the left-right time difference τ, When the left-right time difference τ indicates that neither channel is advanced, it is decided to directly use the downmix signal in the left channel subtraction gain estimation step, the left channel signal subtraction step, the right channel subtraction gain estimation step, and the right channel signal subtraction step, In the left channel subtraction gain estimation step, the left channel signal subtraction step, the right channel subtraction gain estimation step and the right channel signal subtraction step, The downmix signal obtained in the downmixing step is replaced with the downmix signal or the delayed downmix signal determined in the time shifting step.

9. A method for decoding a sound signal, which decodes an input code frame by frame to obtain a sound signal, characterized in that: include: A mono decoding step, decoding the input mono coded CM and obtaining a mono decoded sound signal; A stereo decoding step, decoding the input stereo code CS and obtaining a left channel decoded differential signal and a right channel decoded differential signal; A left channel subtraction gain decoding step, decoding the input left channel subtraction gain code Cα and obtaining a left channel subtraction gain α; The left channel signal addition step is to add the sample value y^y of the left channel decoded differential signal according to each corresponding sample t. L (t), and the sampling value of the mono decoded sound signal ^ x M (t) and the value obtained by adding the value multiplied by the left channel subtraction gain α^y L (t)+α×^x M (t) is obtained as the decoded sound signal of the left channel; A right channel subtraction gain decoding step, decoding the input right channel subtraction gain code Cβ and obtaining a right channel subtraction gain β; The right channel signal addition step is to add the sampling value y^y of the right channel decoded differential signal according to each corresponding sample t. R (t), and the sampling value of the mono decoded sound signal ^ x M (t) and the value obtained by adding the value multiplied by the right channel subtraction gain β^y R (t)+β×^x M The sequence of (t) is obtained as the right channel decoded sound signal, The number of bits used for decoding the mono decoded audio signal in the mono decoding step is set to b M , the number of bits used for decoding the left channel decoded differential signal in the stereo decoding step is set to b L , the number of bits used for decoding the right channel decoded differential signal in the stereo decoding step is set to b R , In the left channel subtraction gain decoding step, Decode the left channel subtraction gain code Cα and obtain a decoded value ^r L , Get the left channel correction coefficient c L and the decoded value obtained by decoding the left channel subtraction gain code Cα L The multiplication value is used as the left channel subtraction gain α, and the left channel correction coefficient c L is a value greater than 0 and less than 1, and is in b L =b M When 0.5, b L B M The more it is, the closer it is to 0 than 0.

5. L B M The smaller the value, the closer it is to 1 than 0.

5. In the right channel subtraction gain decoding step, Decode the right channel subtraction gain code Cβ and obtain a decoded value ^r R , Get the right channel correction coefficient c R and the decoded value obtained by decoding the right channel subtraction gain code Cβ R The multiplication value is used as the right channel subtraction gain β, and the right channel correction coefficient c R is a value greater than 0 and less than 1, and is in b R =b M When 0.5, b R B M The more it is, the closer it is to 0 than 0.

5. R B M The smaller the value, the closer it is to 1 than 0.

5.

10. The sound signal decoding method according to claim 9, characterized in that: Set the number of samples per frame to T, The left channel correction coefficient c L =[Mathematical formula 33] The right channel correction coefficient c R =[Mathematical formula 34] 11. The sound signal decoding method according to claim 9 or 10, characterized in that: Also includes: The left-right time difference decoding step obtains the left-right time difference τ from the input left-right time difference code Cτ; as well as a time shifting step, in which, when the left-right time difference τ indicates that the left channel is ahead, it is decided to directly use the mono decoded sound signal in the left channel signal adding step, and it is decided to use a delayed mono decoded signal in the right channel signal adding step, wherein the delayed mono decoded signal is a signal that delays the mono decoded sound signal by a magnitude corresponding to the magnitude indicated by the left-right time difference τ; in which, when the left-right time difference τ indicates that the right channel is ahead, it is decided to directly use the mono decoded sound signal in the right channel signal adding step, and it is decided to use a delayed mono decoded signal in the left channel signal adding step, wherein the delayed mono decoded signal is a signal that delays the mono decoded sound signal by a magnitude corresponding to the magnitude indicated by the left-right time difference τ; When the left-right time difference τ indicates that neither channel is in advance, it is determined to directly use the monaural decoded audio signal in the left channel signal addition step and the right channel signal addition step, In the left channel signal adding step and the right channel signal adding step, The monaural decoded sound signal or the delayed monaural decoded sound signal determined in the time shift step is used instead of the monaural decoded sound signal obtained in the monaural decoding step.

12. A sound signal encoding device, which encodes an input sound signal for each frame, characterized in that: include: A downmixing unit that obtains a downmixed signal that is a signal obtained by mixing the input left channel input sound signal and the input right channel input sound signal; A left channel subtraction gain estimation unit, which obtains a left channel subtraction gain α and a code representing the left channel subtraction gain α, namely, a left channel subtraction gain code Cα, according to the left channel input sound signal and the downmix signal; The left channel signal subtraction unit obtains the sample value x of the left channel input sound signal for each corresponding sample t. L (t) by subtracting the sample value x of the downmix signal M (t) The value x obtained by multiplying the value of the left channel subtraction gain α L (t)-α×x M The sequence of (t) is taken as the left channel difference signal; a right channel subtraction gain estimation unit, which obtains a right channel subtraction gain β and a code representing the right channel subtraction gain β, namely, a right channel subtraction gain code Cβ, according to the right channel input sound signal and the downmix signal; The right channel signal subtraction unit obtains the sampling value x of the sound signal input from the right channel for each corresponding sample t. R (t) by subtracting the sample value x of the downmix signal M The value x obtained by multiplying (t) and the right channel subtraction gain β R (t)-β×x M The sequence of (t) is used as the right channel difference signal; a mono encoding unit, encoding the downmix signal to obtain a mono code CM; and A stereo encoding unit encodes the left channel difference signal and the right channel difference signal to obtain a stereo code CS, The number of bits used for encoding the downmix signal in the monaural encoding unit is set to b M , the number of bits used for encoding the left channel difference signal in the stereo encoding unit is set to b L , the number of bits used for encoding the right channel difference signal in the stereo encoding unit is set to b R , The left channel subtraction gain estimation unit obtains the left channel correction coefficient c L and the normalized inner product value r of the downmix signal for the left channel input sound signal L The quantized value of the multiplication value is used as the left channel subtraction gain α, and the inner product value r with the left channel subtraction gain α or the normalized inner product value r is obtained. L The quantized value corresponding to the encoding is used as the left channel subtraction gain encoding Cα, the left channel correction coefficient c L is a value greater than 0 and less than 1, and is in b L =b M When 0.5, b L B M The more it is, the closer it is to 0 than 0.

5. L B M The smaller the value, the closer it is to 1 than 0.

5. The right channel subtraction gain estimation unit obtains the right channel correction coefficient c R and the normalized inner product value r of the downmix signal for the right channel input sound signal R The quantized value of the multiplication value is used as the right channel subtraction gain β, and the inner product value r with the right channel subtraction gain β or the normalized inner product value r is obtained. R The quantized value corresponding to the encoding is used as the right channel subtraction gain encoding Cβ, the right channel correction coefficient c R is a value greater than 0 and less than 1, and is in b R =b M When 0.5, b R B M The more it is, the closer it is to 0 than 0.

5. R B M The smaller the value, the closer it is to 1 than 0.

5.

13. A sound signal encoding device, which encodes an input sound signal for each frame, characterized in that: include: A downmixing unit that obtains a downmixed signal that is a signal obtained by mixing the input left channel input sound signal and the input right channel input sound signal; A left channel subtraction gain estimation unit, which obtains a left channel subtraction gain α and a code representing the left channel subtraction gain α, namely, a left channel subtraction gain code Cα, according to the left channel input sound signal and the downmix signal; The left channel signal subtraction unit obtains the sample value x of the left channel input sound signal for each corresponding sample t. L (t) by subtracting the sample value x of the downmix signal M (t) The value x obtained by multiplying the value of the left channel subtraction gain α L (t)-α×x M The sequence of (t) is taken as the left channel difference signal; a right channel subtraction gain estimation unit, which obtains a right channel subtraction gain β and a code representing the right channel subtraction gain β, namely, a right channel subtraction gain code Cβ, according to the right channel input sound signal and the downmix signal; The right channel signal subtraction unit obtains the sampling value x of the sound signal input from the right channel for each corresponding sample t. R (t) by subtracting the sample value x of the downmix signal M The value x obtained by multiplying (t) and the right channel subtraction gain β R (t)-β×x M The sequence of (t) is used as the right channel difference signal; a mono encoding unit, encoding the downmix signal to obtain a mono code CM; and A stereo encoding unit encodes the left channel difference signal and the right channel difference signal to obtain a stereo code CS, The number of bits used for encoding the downmix signal in the monaural encoding unit is set to b M , the number of bits used for encoding the left channel difference signal in the stereo encoding unit is set to b L , the number of bits used for encoding the right channel difference signal in the stereo encoding unit is set to b R , The left channel subtraction gain estimation unit obtains the left channel correction coefficient c L , a normalized inner product value r of the downmix signal for the left channel input sound signal L , and a pre-set value greater than 0 and less than 1, that is, a quantized value of the multiplication value of the left channel coefficient value as the left channel subtraction gain α, and obtain the inner product value r with the left channel subtraction gain α and the normalized inner product value r L The quantized value, or the normalized inner product value r L The code corresponding to the quantized value of the value multiplied by the left channel coefficient value is used as the left channel subtraction gain code Cα, and the left channel correction coefficient c L is a value greater than 0 and less than 1, and is in b L =b M When 0.5, b L B M The more it is, the closer it is to 0 than 0.

5. L B M The smaller the value, the closer it is to 1 than 0.

5. The right channel subtraction gain estimation unit obtains the right channel correction coefficient c R , a normalized inner product value r of the downmix signal for the right channel input sound signal R , and a preset value greater than 0 and less than 1, that is, a quantized value of the multiplication value of the right channel coefficient value as the right channel subtraction gain β, and obtain the inner product value r with the right channel subtraction gain β and the normalized inner product value r R The quantized value or the normalized inner product value r R The code corresponding to the quantized value of the value multiplied by the left channel coefficient value is used as the right channel subtraction gain code Cβ, and the right channel correction coefficient c R is a value greater than 0 and less than 1, and is in b R =b M When 0.5, b R B M The more it is, the closer it is to 0 than 0.

5. R B M The smaller the value, the closer it is to 1 than 0.

5.

14. A sound signal encoding device, which encodes an input sound signal for each frame, characterized in that: include: A downmixing unit that obtains a downmixed signal that is a signal obtained by mixing the input left channel input sound signal and the input right channel input sound signal; A left channel subtraction gain estimation unit, which obtains a left channel subtraction gain α and a code representing the left channel subtraction gain α, namely, a left channel subtraction gain code Cα, according to the left channel input sound signal and the downmix signal; The left channel signal subtraction unit obtains the sample value x of the left channel input sound signal for each corresponding sample t. L (t) by subtracting the sample value x of the downmix signal M (t) The value x obtained by multiplying the value of the left channel subtraction gain α L (t)-α×x M The sequence of (t) is taken as the left channel difference signal; a right channel subtraction gain estimation unit, which obtains a right channel subtraction gain β and a code representing the right channel subtraction gain β, namely, a right channel subtraction gain code Cβ, according to the right channel input sound signal and the downmix signal; The right channel signal subtraction unit obtains the sampling value x of the sound signal input from the right channel for each corresponding sample t. R (t) by subtracting the sample value x of the downmix signal M The value x obtained by multiplying (t) and the right channel subtraction gain β R (t)-β×x M The sequence of (t) is used as the right channel difference signal; a mono encoding unit, encoding the downmix signal to obtain a mono code CM; and A stereo encoding unit encodes the left channel difference signal and the right channel difference signal to obtain a stereo code CS, The number of bits used for encoding the downmix signal in the monaural encoding unit is set to b M , the number of bits used for encoding the left channel difference signal in the stereo encoding unit is set to b L , the number of bits used for encoding the right channel difference signal in the stereo encoding unit is set to b R , The left channel subtraction gain estimation unit obtains the left channel correction coefficient c L , a normalized inner product value r of the downmix signal for the left channel input sound signal L , the quantized value of the multiplication value of the left channel coefficient value determined for each frame, which is greater than or equal to 0 and less than or equal to 1, is used as the left channel subtraction gain α, and the normalized inner product value r is obtained. L The quantized value, or the normalized inner product value r L The code corresponding to the quantized value of the value multiplied by the left channel coefficient value is used as the left channel subtraction gain code Cα, and the left channel correction coefficient c L is a value greater than 0 and less than 1, and is in b L =b M When 0.5, b L B M The more it is, the closer it is to 0 than 0.

5. L B M The smaller the value, the closer it is to 1 than 0.

5. The right channel subtraction gain estimation unit obtains the right channel correction coefficient c R , a normalized inner product value r of the downmix signal for the right channel input sound signal R , the quantized value of the multiplication value of the coefficient value for the right channel determined for each frame, which is greater than or equal to 0 and less than or equal to 1, is used as the right channel subtraction gain β, and the normalized inner product value r is obtained with the right channel subtraction gain β and the normalized inner product value r R The quantized value, or the normalized inner product value r R The code corresponding to the quantized value of the value multiplied by the coefficient value of the right channel is used as the right channel subtraction gain code Cβ, and the right channel correction coefficient c R is a value greater than 0 and less than 1, and is in b R =b M When 0.5, b R B M The more it is, the closer it is to 0 than 0.

5. R B M The smaller the value, the closer it is to 1 than 0.

5.

15. The sound signal encoding device according to claim 14, characterized in that: The invention further includes a left-right correlation estimation unit, which obtains a left-right correlation coefficient as a correlation coefficient between the left channel input sound signal and the right channel input sound signal. The left channel subtraction gain estimation unit uses the left-right correlation coefficient as the left channel coefficient value. The right channel subtraction gain estimation unit uses the left-right correlation coefficient as the coefficient value for the right channel.

16. The sound signal encoding device according to any one of claims 12 to 14, characterized in that: The system further includes a left-right relationship information estimation unit, which obtains information indicating which channel of the left channel and the right channel precedes, namely, preceding channel information, and a correlation coefficient between the left channel input sound signal and the right channel input sound signal, namely, a left-right correlation coefficient. The downmixing unit performs weighted averaging of the left channel input sound signal and the right channel input sound signal based on the preceding channel information and the left-right correlation coefficient, and obtains the downmixed signal in a manner such that the larger the left-right correlation coefficient, the greater the inclusion of the preceding channel input sound signal in the left channel input sound signal and the right channel input sound signal.

17. The sound signal encoding device according to any one of claims 12 to 14, characterized in that: Set the number of samples per frame to T, The left channel correction coefficient c L =[Mathematical formula 35] The right channel correction coefficient c R =[Mathematical formula 36] 18. The sound signal encoding device according to any one of claims 12 to 14, characterized in that: ε L , ε R , ε M are values ​​greater than 0 and less than 1 respectively. The left channel subtraction gain estimation unit uses the inner product value E of the left channel input sound signal, the downmix signal, and the previous frame. L (-1), through [Mathematical formula 37] Get the inner product value E L (0), Use the energy E of the downmix signal and the downmix signal of the previous frame M (-1), through [Mathematical formula 38] Get the energy E of the downmix signal M (0), The inner product value E will be used L (0) and the energy E of the downmix signal M (0), through [Mathematical formula 39] r L =And L (0) / E M (0) The obtained r L The inner product value normalized for the left channel input sound signal as the downmix signal, The right channel subtraction gain estimation unit uses the right channel input sound signal, the downmix signal, and the inner product value E of the previous frame. R (-1), through [Math 40] Get the inner product value E L (0), Use the energy E of the downmix signal and the downmix signal of the previous frame M (-1), By [Mathematical formula 41] Get the energy E of the downmix signal M (0), The inner product value E will be used L (0) and the energy E of the downmix signal M (0), through [Mathematical formula 42] r R =And R (0) / E M (0) The obtained r R The inner product value normalized with respect to the right channel input sound signal is used as the downmix signal.

19. The sound signal encoding device according to any one of claims 12 to 14, characterized in that: Also includes: A left-right time difference estimation unit, which obtains a left-right time difference τ and a code representing the left-right time difference τ, that is, a left-right time difference code Cτ, according to the left channel input sound signal and the right channel input sound signal; and The time shift unit determines to use the downmix signal directly in the left channel subtraction gain estimation unit and the left channel signal subtraction unit, and determines to use a delayed downmix signal in the right channel subtraction gain estimation unit and the right channel signal subtraction unit, when the left-right time difference τ indicates that the left channel is ahead, wherein the delayed downmix signal is a signal in which the downmix signal is delayed by a magnitude corresponding to the left-right time difference τ. When the left-right time difference τ indicates that the right channel is ahead, it is determined to directly use the downmix signal in the right channel subtraction gain estimation unit and the right channel signal subtraction unit, and it is determined to use a delayed downmix signal in the left channel subtraction gain estimation unit and the left channel signal subtraction unit, the delayed downmix signal being a signal in which the downmix signal is delayed by a magnitude corresponding to the left-right time difference τ, When the left-right time difference τ indicates that neither channel is advanced, it is determined to directly use the downmix signal in the left channel subtraction gain estimation unit, the left channel signal subtraction unit, the right channel subtraction gain estimation unit, and the right channel signal subtraction unit, The left channel subtraction gain estimation unit, the left channel signal subtraction unit, the right channel subtraction gain estimation unit, and the right channel signal subtraction unit use the downmix signal or the delayed downmix signal determined in the time shift unit instead of the downmix signal obtained in the downmix unit.

20. A sound signal decoding device, which decodes input code for each frame to obtain a sound signal, characterized in that: include: A mono decoding unit, which decodes the input mono coded CM and obtains a mono decoded sound signal; A stereo decoding unit, which decodes the input stereo code CS and obtains a left channel decoded differential signal and a right channel decoded differential signal; A left channel subtraction gain decoding unit, which decodes the input left channel subtraction gain code Cα and obtains the left channel subtraction gain α; The left channel signal adding unit adds the sample value y of the left channel decoded differential signal to the corresponding sample t. L (t), and the sampling value of the mono decoded sound signal ^ x M (t) and the value obtained by adding the value multiplied by the left channel subtraction gain α^y L (t)+α×^x M (t) is obtained as the decoded sound signal of the left channel; A right channel subtraction gain decoding unit, which decodes the input right channel subtraction gain code Cβ and obtains the right channel subtraction gain β; The right channel signal adding unit adds the sampling value of the right channel decoded differential signal to the corresponding sample t. R (t), and the sampling value of the mono decoded sound signal ^ x M (t) and the value obtained by adding the value multiplied by the right channel subtraction gain β^y R (t)+β×^x M The sequence of (t) is obtained as the right channel decoded sound signal, The number of bits used for decoding the monaural decoded audio signal in the monaural decoding unit is set to b M , the number of bits used for decoding the left channel decoding difference signal in the stereo decoding unit is set to b L , the number of bits used for decoding the right channel decoding difference signal in the stereo decoding unit is set to b R , The left channel subtraction gain decoding unit performs: Decode the left channel subtraction gain code Cα and obtain a decoded value ^r L , Get the left channel correction coefficient c L and the decoded value obtained by decoding the left channel subtraction gain code Cα L The multiplication value is used as the left channel subtraction gain α, and the left channel correction coefficient c L is a value greater than 0 and less than 1, and is in b L =b M When 0.5, b L B M The more it is, the closer it is to 0 than 0.

5. L B M The smaller the value, the closer it is to 1 than 0.

5. The right channel subtraction gain decoding unit performs: Decode the right channel subtraction gain code Cβ and obtain a decoded value ^r R , Get the right channel correction coefficient c R and the decoded value obtained by decoding the right channel subtraction gain code Cβ R The multiplication value is used as the right channel subtraction gain β, and the right channel correction coefficient c R is a value greater than 0 and less than 1, and is in b R =b M When 0.5, b R B M The more it is, the closer it is to 0 than 0.

5. R B M The smaller the value, the closer it is to 1 than 0.

5.

21. The sound signal decoding device according to claim 20, characterized in that: Set the number of samples per frame to T, The left channel correction coefficient c L =[Mathematical formula 43] The right channel correction coefficient c R =[Mathematical formula 44] 22. The sound signal decoding device according to claim 20 or 21, characterized in that: Also includes: A left-right time difference decoding unit, which obtains a left-right time difference τ from an input left-right time difference code Cτ; and The time shift unit determines to use the monaural decoded audio signal directly in the left channel signal addition unit, and determines to use a delayed monaural decoded signal in the right channel signal addition unit, when the left-right time difference τ indicates that the left channel is ahead, wherein the delayed monaural decoded signal is a signal in which the monaural decoded audio signal is delayed by a magnitude corresponding to the left-right time difference τ. When the left-right time difference τ indicates that the right channel is ahead, it is determined that the monaural decoded sound signal is directly used in the right channel signal addition unit, and it is determined that a delayed monaural decoded signal is used in the left channel signal addition unit, wherein the delayed monaural decoded signal is a signal in which the monaural decoded sound signal is delayed by a magnitude corresponding to the left-right time difference τ. When the left-right time difference τ indicates that neither channel is in advance, it is determined to directly use the monaural decoded audio signal in the left channel signal adding unit and the right channel signal adding unit, The left channel signal adding section and the right channel signal adding section use the monaural decoded audio signal or the delayed monaural decoded audio signal determined in the time shift section instead of the monaural decoded audio signal obtained in the monaural decoding section.

23. A computer program product, comprising a computer program for causing a computer to execute the steps of the encoding method according to any one of claims 1 to 3.

24. A computer program product comprising a computer program for causing a computer to execute the steps of the decoding method according to claim 9.

25. A computer-readable recording medium recording a program for causing a computer to execute each step of the encoding method according to any one of claims 1 to 3.

26. A computer-readable recording medium recording a program for causing a computer to execute each step of the decoding method according to claim 9.

Images