Processing device, processing method, and processing program

Abstract

To be able to appropriately change a dimension of a line spectrum pair.SOLUTION: A processing unit using line spectral pairs comprises: an LSP inverse transform section 27 (27-m, 27-m+1, ...) that transforms a first line spectral pair in a first dimension into a first linear prediction coefficient in the first dimension; an LSP dimensional transform section 28 (28-m, 28-m+1, ...) that transforms the first linear prediction coefficient in the first dimension into a second linear prediction coefficient in a second dimension; and an LSP dimensional transformation section 29 (29-m, 29-m+1, ...) that transforms the second linear prediction coefficient in the second dimension into a second line spectrum pair in the second dimension.SELECTED DRAWING: Figure 2

Description

[Technical field]

[0001] The present invention relates to a technique that utilizes line spectrum pairs. [Background technology]

[0002] For example, line spectral pairs (LSPs) are used as parameters representing the spectral characteristics of an audio signal (see, for example, Non-Patent Document 1). Compared to other representation methods, line spectral pairs have advantageous features for audio signal compression, such as the ability to perform parameter quantization and time-domain interpolation well. [Prior art documents] [Non-patent literature]

[0003] [Non-Patent Document 1] Fumitada Itakura, "Statistical and mathematical methods for speech signals, from maximum likelihood spectrum method through PARCOR to LSP," Institute of Electronics, Information and Communication Engineers, Fundamentals Review, vol. 3, no. 3, pp. 2-30, Jan. 2010. Summary of the Invention [Problem to be solved by the invention]

[0004] For example, spectral characteristics of a plurality of audio signals are compared using spectral line pairs. However, when the orders of the spectral line pairs of the audio signals to be compared are different, the comparison cannot be performed.

[0005] For example, it is possible to take differences between multiple spectral line pairs and encode them to reduce the amount of data, but if the orders of the spectral line pairs are different, it is not possible to appropriately take differences and encode them. In addition, if the orders of multiple spectral line pairs are made the same in order to take differences between the spectral line pairs, if the order of the spectral line pairs is made high, a signal that can be adequately represented by a low-order spectral line pair must be made into a high-order spectral line pair, which increases the amount of data. On the other hand, if the spectral line pairs are made low, a signal that requires a high-order line spectrum cannot be appropriately represented, resulting in poor quality.

[0006] The present invention has been made in consideration of the above circumstances, and an object of the present invention is to provide a technique capable of appropriately changing the dimension of a line spectrum pair. [Means for solving the problem]

[0007] In order to achieve the above-mentioned object, a processing device according to one aspect is a processing device that uses a line spectral pair, and includes a first linear prediction coefficient conversion unit that converts a first-dimensional first line spectral pair into a first-dimensional first linear prediction coefficient, a linear prediction coefficient dimension conversion unit that converts the first linear prediction coefficient of the first dimension into a second linear prediction coefficient of a second dimension, and a first line spectral pair conversion unit that converts the second linear prediction coefficient of the second dimension into a second line spectral pair of the second dimension. Effect of the Invention

[0008] According to the present invention, the dimension of the line spectrum pair can be appropriately changed. [Brief description of the drawings]

[0009] [Figure 1] FIG. 1 is a diagram showing the overall configuration of a processing system according to an embodiment. [Diagram 2] FIG. 2 is a functional configuration diagram of an LSP encoding processing unit of an encoder according to an embodiment. [Diagram 3]FIG. 3 is a functional configuration diagram of an LSP decoding processing unit of a decoder according to an embodiment. [Figure 4] FIG. 4 is a flowchart of an encoding process according to an embodiment. [Diagram 5] FIG. 5 is a flowchart of a decoding process according to an embodiment. [Figure 6] FIG. 6 is a diagram showing the relationship between the order and the coefficients corresponding to each dimension when an LSP is calculated from PCM data according to the comparative example. [Figure 7] FIG. 7 is a diagram showing the relationship between the dimension and the coefficients corresponding to each dimension when the dimension of the LSP is increased by the dimension conversion process according to an embodiment. [Figure 8] FIG. 8 is a diagram showing the relationship between dimensions and coefficients corresponding to each dimension when the dimensions of an LSP are reduced by a dimension conversion process according to an embodiment. [Figure 9] FIG. 9 is a diagram illustrating a configuration of a computer device according to an embodiment. [Figure 10] FIG. 10 is a functional configuration diagram of an LSP encoding processing unit of an encoder according to a modified example. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0010] The following embodiments will be described with reference to the drawings. Note that the following embodiments do not limit the scope of the invention, and not all of the elements and combinations thereof described in the embodiments are necessarily essential to the solution of the invention.

[0011] First, a processing system according to an embodiment will be described.

[0012] FIG. 1 is a diagram showing the overall configuration of a processing system according to an embodiment.

[0013] The processing system 1 includes an encoder 10 as an example of a processing device, and a decoder 40 as an example of a processing device. The encoder 10 and the decoder 40 are connected via a network 60. The network 60 is, for example, a Local Area Network (LAN) or a Wide Area Network (WAN).

[0014] The encoder 10 includes a target data input unit 11, an LSP encoding processing unit 12, a target data encoding unit 13, and a compressed data output unit 14.

[0015] The target data input unit 11 inputs data to be compressed (encoded). The target data may be, for example, time-series audio data converted by PCM (Pulse Code Modulation). The target data may be obtained from the auxiliary storage device 106 (see FIG. 9) in the encoder 10, or from a recording device (not shown).

[0016] The LSP encoding processing unit 12 creates linear prediction coefficients (LPC) to be applied to data sequences (segment data sequences) of each segment (time segment) of time-series voice data, and executes an encoding process to create encoded data (LSP encoded data) that encodes the differences between segments for line spectral pairs (LSPs) corresponding to the LPC.

[0017] The target data encoding unit 13 executes an encoding process in which each section data string of the voice data is encoded using an LPC corresponding to each section to generate encoded data (section encoded data: first section encoded data, second section encoded data, etc.).

[0018] The compressed data output unit 14 creates compressed data including the LSP encoded data for each section created by the LSP encoding processing unit 12 and the section encoded data for each section encoded by the target data encoding unit 13, and outputs the data to the decoder 40. In this embodiment, the compressed data includes information indicating the dimension of the LSP corresponding to each section included in each LSP encoded data.

[0019] The decoder 40 includes a compressed data input unit 41, an LSP decoding processing unit 42, a target data calculation unit 43, and a target data processing unit 44. Here, the compressed data input unit 41 is an example of a compressed data receiving unit, and the target data calculation unit 43 is an example of a restoration unit.

[0020] The compressed data input unit 41 inputs compressed data (data to be processed) transmitted from the encoder 10 via the network 60.

[0021] The LSP decoding processing unit 42 executes a decoding process in which, based on the LSP encoded data included in the compressed data, the LSP corresponding to each section of the voice data is decoded and the LPC corresponding to the decoded LSP is output to the target data calculation unit 43.

[0022] The target data calculation unit 43 calculates a predicted value for each time point of each section based on the LPC corresponding to each section received from the LSP decoding processing unit 42, and performs a restoration process to restore the original voice data (compression target data) by adding this predicted value to the prediction error for each time point obtained by decoding the section encoded data of each section.

[0023] The target data processing unit 44 processes the restored audio data to output audio.

[0024] Next, the LSP encoding unit 12 of the encoder 10 will be described in detail.

[0025] Fig. 2 is a functional block diagram of an LSP encoding processing unit of an encoder according to an embodiment. For convenience, Fig. 2 shows functional units used to process a section data string xm[t] (m is an arbitrary integer: an example of a first section) of section m among multiple sections of data to be compressed, and a section data string xm+1[t] (an example of a second section) of section m+1.

[0026] The LSP encoding processing unit 12 has an LPC analysis unit 20 (20-m, 20-m+1, etc.), an LSP conversion unit 21 (21-m, 21-m+1, etc.), a difference calculation unit 22 (22-m, 22-m+1, etc.), an LSP quantization unit 23 (23-m, 23-m+1, etc.), an LSP inverse quantization unit 24 (24-m, 24-m+1, etc.), an addition unit 25 (25-m, 25-m+1, etc.), an LSP dimension conversion unit 26 (26-m, 26-m+1, etc.), and an LSP encoding unit 30 (30-m, 30-m+1, etc.) corresponding to each section of the compression target data. Note that each functional unit only needs to be present when the processing is executed, and does not always have to be present.

[0027] The LPC analysis unit 20 acquires a section data sequence (first section data sequence, second section data sequence, ...) of a corresponding section (target section) from the target data input unit 11, and performs an analysis process to analyze the dimension of an LPC suitable for the target section based on the section data sequence, and to analyze the coefficient of each dimension of the LPC. For example, LPCs of multiple orders may be obtained to determine an appropriate order. Note that a known technique can be used for the analysis process. The LPC analysis unit 20 outputs the LPC (first linear prediction coefficient, an example of a first linear prediction coefficient corresponding coefficient) obtained in the analysis process to the target data encoding unit 13, and also outputs it to the LSP conversion unit 21. Here, when xm+1[t] is the second section, the LPC analysis unit 20-m+1 corresponds to a specification unit, and the output LPC corresponds to a third linear prediction coefficient (an example of a third linear prediction coefficient corresponding coefficient).

[0028] The LSP conversion unit 21 converts the LPC output from the LPC analysis unit 20 into an LSP of a predetermined order ns, and outputs it to the difference calculation unit 22. Here, the predetermined order ns may be determined by evaluating LSPs of a plurality of orders. For example, the predetermined order may be determined by selecting a method according to the purpose, such as using a criterion based on AIC (Akaike Information Criterion) or MDL (minimum description length). Note that a known technique can be used for the process of converting an LPC into an LSP. The LSP conversion unit 21 outputs the order ns to the target data encoding unit 13. Here, when xm+1[t] is the second section, the LSP conversion unit 21-m+1 corresponds to the second line spectrum pair conversion unit, and the output LSP corresponds to the third line spectrum pair.

[0029] The difference calculation unit 22 calculates difference data between the LSP output from the LSP conversion unit 21 and the LSP output from the LSP conversion unit 29 of the previous section (m-1 when the target section is m) (29-m-1 when the target section is m-1), and outputs the difference data to the LSP quantization unit 23. Here, the difference data is data including differences for each dimension of the LSP, and is data of the same dimension as the LSP. Note that when the target section is the first section, since the LSP conversion unit 29 of the previous section does not exist, a default value (e.g., zero) may be used as the output value of the LSP conversion unit 29 of the previous section.

[0030] The LSP quantizer quantizes the difference data output from the difference calculator 22, and outputs the quantized difference data (quantized difference data: difference corresponding data) to the LSP encoder 30 and the LSP inverse quantizer .

[0031] The LSP encoding unit 30 outputs the quantized difference data output from the LSP quantization unit 23 to the compressed data output unit 14 as LSP encoded data, either directly or after performing a predetermined encoding process.

[0032] The LSP inverse quantization unit 24 performs an inverse quantization process on the quantized difference data output from the LSP quantization unit 23 to restore the difference data (inverse quantized difference data), and outputs the difference data to the addition unit 25. Note that quantization and inverse quantization include errors, and strictly speaking the data may not be completely identical to the original difference data. However, in cases where there is no need to take the errors into consideration and there is no problem in treating the data as the same, the data will be described as difference data for convenience.

[0033] The adder 25 calculates an LSP corresponding to the LPC of the target section by adding the difference data output from the LSP inverse quantizer 24 and the LSP output from the LSP converter 29 of the previous section (m-1 when the target section is m), and outputs the LSP to the LSP dimension converter 26. Note that the LSP obtained by the adder 25 may not be completely identical to the LSP output from the LSP converter 21, but when there is no problem in treating them as the same, they will be described as the LSP of the target section for convenience. Note that when the target section is the first section, the LSP converter 29 of the previous section does not exist, so a default value (for example, zero) may be used as the output value of the LSP converter 29 of the previous section.

[0034] The LSP dimension conversion unit 26 performs a dimension conversion process to convert the LSP in a target section (eg, m) so that the dimension becomes the same as the dimension of the LSP in the next section (eg, m+1).

[0035] LSP dimension conversion units 26 (26-m, 26-m+1, etc.) include LSP inverse conversion units 27 (27-m, 27-m+1, etc.), LPC dimension conversion units 28 (28-m, 28-m+1, etc.), and LSP conversion units 29 (29-m, 29-m+1, etc.).

[0036] The LSP inverse transform unit 27 is an example of a first linear prediction coefficient transform unit, and transforms the LSP (first line spectrum pair) of a predetermined dimension (first dimension) output from the adder 25 into an LPC of the same dimension and outputs it to the LPC dimension transform unit .

[0037] The LPC dimension conversion unit 28 is an example of a linear prediction coefficient dimension conversion unit, which converts the LPC output from the LSP inverse conversion unit 27 to the same dimension as the LPC in the next section, and outputs the converted LPC to the LSP conversion unit 29. Here, for the current dimension, it may be determined from the configuration of the LPC, or the dimension information may be obtained from the LSP conversion unit 21 or the like. Also, for the dimension of the LPC in the next section, it may be obtained from the LSP conversion unit 21 or the like in the next section.

[0038] Next, the LPC dimension conversion process for converting the dimension of the LPC by the LPC dimension conversion unit 28 will be described in detail. Here, let the number of dimensions of the source LPC be ns, the number of dimensions of the destination LPC be nd, and the coefficients of each dimension of the LPC be a1, a2, a3,... in order from the first order.

[0039] First, the case where the order of the LPC is increased (when ns < nd) will be described.

[0040] The source LPC is {a1, a2, a3,..., a ns}, and for the coefficients of orders higher than ns of the LPC after conversion, since they are coefficients representing the influence degree of data farther from the data at the time point t to be predicted, assuming the influence is weak, for example, they are set to 0. That is, for the coefficients of dimensions greater than the number of dimensions ns and up to the number of dimensions nd after conversion, they are each set to 0. Therefore, the LPC dimension conversion unit 28 sets the LPC after conversion to {a1, a2, a3,..., a ns ..., 0, 0, 0}.

[0041] Next, the case where the order of the LPC is decreased (when ns > nd) will be described.

[0042] First, for the source LPC {a1 ns , a2 ns , a3 ns ,..., a ns ns}, using the following formula (1), the LPC of order ns - 1 with one dimension lower {a1 ns-1 , a2 ns-1 , a3ns-1 ,···,a ns-1 ns-1} is calculated (predicted). Here, a i The value on the right hand side of (e.g., ns, ns-1) represents the order of the LPC to which the coefficient corresponds. For example, a1 ns indicates the first-order coefficient of the ns-order LPC, and a1 ns-1 indicates the first-order coefficient of the ns-1-th order LPC.

[0043]

number

[0044] Next, by replacing equation (1) with the LPC of the target order, the process of calculating the LPC of one dimension lower (prediction process) is performed on the nd-th order LPC {a1 nd ,a2 nd ,a3 nd ,···,a nd nd} is calculated.

[0045] Next, the derivation of equation (1) will be explained.

[0046] Here, the data string x(t) to be processed can be expressed by an ns-order autoregressive model shown in equation (2).

[0047]

number

[0048] Here, ε ns (t) denotes the prediction error, and a k ns indicates the k-th order coefficient (prediction coefficient) in the ns-th order LPC.

[0049] Here, the autocorrelation coefficient of the data sequence x(t) is r t Then, assume that the ns-th order equation can be expressed as shown in the following equation (3).

[0050]

number

[0051] Due to the symmetry of the coefficient matrix, equation (3) can be transformed into the following equation (4).

[0052]

number

[0053] If equation (4) is transformed so as to delete the top line, it becomes as shown in equation (5) below.

[0054]

number

[0055] Moreover, when the bottom line of equation (3) is deleted, the equation becomes as shown in equation (6) below.

[0056]

number

[0057] Also, the autocorrelation coefficient of the data sequence x(t) is r t In this case, the ns-1 order equation can be expressed as shown in the following equation (7).

[0058]

number

[0059] By transforming equation (7) and substituting it into equation (5), it can be expressed as shown in equation (8).

[0060]

number

[0061] Moreover, by transforming equation (7) and substituting it into equation (6), it can be expressed as shown in equation (9).

[0062]

number

[0063] Focusing on the first term in equation (8), we substitute it into equation (9) and eliminate the matrix containing r to obtain equation (10).

[0064]

number

[0065] By rearranging this equation (10), the above equation (1) is obtained.

[0066] The LSP conversion unit 29 is an example of a first line spectrum pair conversion unit, and converts the dimension-converted LPC output by the LPC dimension conversion unit 28 into an LSP of the same dimension (an example of a second line spectrum pair) and outputs it to the difference calculation unit 22 for the next section and the adder unit 25. Here, since an LSP of the same dimension as the LPC (LSP) of the next section is transmitted to the difference calculation unit 22 and the adder unit 25 for the next section, calculation processing can be performed directly in the difference calculation unit 22 and the adder unit 25. Furthermore, since the difference of the LSP can be calculated, the data amount of information related to the LSP can be reduced, and the data amount transmitted to the decoder 40 can be reduced.

[0067] Next, the LSP decoding unit 42 of the decoder 40 will be described in detail.

[0068] Fig. 3 is a functional block diagram of an LSP decoding processing unit of a decoder according to an embodiment. For convenience, Fig. 3 shows functional units used for processing corresponding to a section data string xm[t] (m is an arbitrary integer) of section m among multiple sections of compression target data and a section data string xm+1[t] of section m+1.

[0069] The LSP decoding unit 42 has, corresponding to each section of the compression target data, an LSP decoding unit 50 (50-m, 50-m+1, etc.), an LSP inverse quantization unit 51 (51-m, 51-m+1, etc.), an addition unit 52 (52-m, 52-m+1, etc.), an LSP inverse transformation unit 53 (53-m, 53-m+1, etc.), and an LSP dimension transformation unit 54 (54-m, 54-m+1, etc.). Note that each functional unit only needs to be present when processing is executed, and does not have to be present all the time.

[0070] The LSP decoding unit 50 obtains the LSP encoded data of the corresponding section (target section) from the compressed data input unit 41, and outputs data related to the LSP indicated by the LSP encoded data (here, LSP difference data) to the LSP inverse quantization unit 51. Note that, when a predetermined encoding process has been performed on the LSP encoded data, the LSP decoding unit 50 may perform a decoding process corresponding to the LSP encoded data.

[0071] The LSP inverse quantization unit 51 performs inverse quantization processing on the LSP difference data output from the LSP decoding unit 50 , and outputs the result to the addition unit 52 .

[0072] The adder 52 calculates an LSP (an example of a first line spectrum pair) corresponding to the LPC of the target section by adding the difference data output from the LSP inverse quantizer 51 and the LSP output from the LSP converter 57 (57-m-1 when the target section is m) of the previous section (when the target section is m, m-1), and outputs the LSP to the LSP inverse converter 53 and the LSP dimension converter 54. Note that when the target section is the first section, the LSP converter 57 of the previous section does not exist, so a default value (e.g., zero) may be used as the output value of the LSP converter 57 of the previous section. Here, the calculation data in the claims is, for example, the difference data of the target section and the LSP from the previous section, but more specifically, it can be, for example, a default value and a set of difference data of all sections prior to the target section. Here, when xm[t] is the first interval and xm+1[t] is the second interval, the adder 52-m is an example of a first line spectrum pair calculation unit, the adder 52-m+1 is an example of a second line spectrum pair conversion unit, and the output LSP is an example of a third line spectrum pair and data corresponding to the third line spectrum pair.

[0073] The LSP inverse transform unit 53 transforms the LSP (first line spectrum pair) of a predetermined dimension (first dimension) output from the adder unit 52 into an LPC of the same dimension, and outputs it to the target data calculation unit 43. Here, when xm[t] is the first interval and xm+1[t] is the second interval, the LSP inverse transform unit 53-m is an example of a first line spectrum pair, the LSP inverse transform unit 53-m+1 is an example of a second linear prediction coefficient transform unit, and the output LPC is an example of a third linear prediction coefficient corresponding coefficient.

[0074] The LSP dimension conversion unit 54 performs a dimension conversion process to convert the LSP in a target section (eg, m) so that the dimension becomes the same as the dimension of the LSP in the next section (eg, m+1).

[0075] LSP dimension conversion units 54 (54-m, 54-m+1, etc.) include an LSP inverse conversion unit 55 (55-m, 55-m+1, etc.), an LPC dimension conversion unit 56 (56-m, 56-m+1, etc.), and an LSP conversion unit 57 (57-m, 57-m+1, etc.).

[0076] The LSP inverse transform unit 55 is an example of a first linear prediction coefficient transform unit, and transforms the LSP (first line spectrum pair) of a predetermined dimension (first dimension) output from the adder 52 into an LPC of the same dimension and outputs it to the LPC dimension transform unit 56.

[0077] The LPC dimension conversion unit 56 is an example of a linear prediction coefficient dimension conversion unit, which performs LPC dimension conversion processing on the LPC output from the LSP inverse conversion unit 55 so that the LPC has the same dimension (second dimension) as the LPC of the next section, and outputs the converted LPC to the LSP conversion unit 57. Here, the current dimension may be determined from the configuration of the LPC, or dimension information transmitted from the encoder 10 may be obtained from the LSP decoding unit 50 or the like. Furthermore, the dimension information of the LPC of the next section may be obtained from the LSP decoding unit 50 or the like of the next section. Note that the dimension conversion processing is similar to the LPC dimension conversion processing of the LPC dimension conversion unit 28 of the encoder 10.

[0078] The LSP converter 57 is an example of a first line spectrum pair converter, and converts the dimension-converted LPC output by the LPC dimension converter 56 into an LSP (second line spectrum pair) of the same dimension, and outputs it to the adder 52 for the next interval. Here, since an LSP of the same dimension as the LPC (LSP) of the next interval is transmitted to the adder 52 for the next interval, the adder 52 can perform calculation processing as is.

[0079] Next, the processing operation in the processing system 1 will be described.

[0080] First, the encoding process by the encoder 10 will be described.

[0081] FIG. 4 is a flowchart of an encoding process according to an embodiment.

[0082] In the encoding process, first, the target data input unit 11 accepts audio data to be encoded (S11).

[0083] Next, the LSP encoding processing unit 12 determines whether or not the section data strings of all sections have been processed (S12), and if the data strings of all sections have not been processed (S12:N), the processing proceeds to step S13 and processing is performed on the data string of the relevant section.

[0084] In step S13, the LSP encoding processing unit 12 determines an LPC to be used for the data sequence in the target section, performs a process of encoding information about the LSP corresponding to the LPC, and outputs data about the encoded LSP (e.g., LSP difference data) to the compressed data output unit 14.

[0085] Next, the target data encoding unit 13 calculates a prediction error between the data value at each time point of this interval and the predicted value by the LPC based on the LPC to be used in each interval determined by the LSP encoding processing unit 12 (S14). Next, the target data encoding unit 13 performs a predetermined encoding process on the calculated prediction error to encode it, outputs the encoded data (interval encoded data) to the compressed data output unit 14 (S15), and the process proceeds to step S12.

[0086] On the other hand, if the data strings of all sections have been processed (S12: Y), the LSP encoding unit 12 advances the process to step S16.

[0087] In step S16, the compressed data output unit 14 creates compressed data for the entire audio data by combining the LSP difference data and the section encoded data for all sections, and outputs (transmits) the compressed data to the decoder 40.

[0088] In this encoding process, even if the dimensions of the LPC in each section are different, the dimensions of the LSPs in the sections can be made the same, making it possible to properly calculate differential data, and thus reducing the amount of data related to the LSPs.

[0089] Next, the decoding process in the decoder 40 will be described.

[0090] FIG. 5 is a flowchart of a decoding process according to an embodiment.

[0091] The compressed data input unit 41 inputs compressed data (an example of data to be processed) transmitted from the encoder 10 via the network 60 (S31).

[0092] Next, the LSP decoding processing unit 42 determines whether or not the data sequences of all sections in the compressed data have been processed (S32), and if not all processing units have been processed (S32:N), the processing proceeds to step S33, and processing is performed on the unprocessed sections.

[0093] In step S33, the LSP decoding unit 42 executes a decoding process in which it decodes the LSP corresponding to each section based on the LSP encoded data included in the compressed data, and outputs the LPC corresponding to the decoded LSP to the target data calculation unit 43.

[0094] Next, the target data calculation unit 43 restores the original voice data (compression target data) based on the LPC corresponding to each section received from the LSP decoding processing unit 42 and each section encoded data (S34), and the process proceeds to step S32.

[0095] On the other hand, if the data strings of all sections have been processed (S32: Y), the LSP decoding unit 42 advances the process to step S35.

[0096] In step S35, the target data calculation unit 43 creates voice data by collecting the restored data for all sections, and the target data processing unit 44 performs voice output processing using the created voice data, and the process ends.

[0097] According to the above-described decoding process, it is possible to appropriately output audio based on the compressed data created by the encoder 10.

[0098] Next, the accuracy of the LSP obtained by converting the dimension through the LSP dimension conversion process according to this embodiment will be described.

[0099] Fig. 6 is a diagram showing the relationship between the order of an LSP and a coefficient corresponding to each dimension when an LSP is calculated from PCM data according to a comparative example. Fig. 7 is a diagram showing the relationship between the dimension and a coefficient corresponding to each dimension when the dimension of an LSP is increased by a dimension conversion process according to an embodiment. Fig. 8 is a diagram showing the relationship between the dimension and a coefficient corresponding to each dimension when the dimension of an LSP is decreased by a dimension conversion process according to an embodiment.

[0100] When the LCP of each dimension is calculated by the least squares method using PCM data of a predetermined musical tone as an example of audio data, and the LSP corresponding to each dimension is calculated, the coefficients of the LSP are as shown in FIG.

[0101] Here, when an LSP of a predetermined dimension calculated using PCM data is increased by one dimension by the LSP dimension conversion process according to this embodiment, each coefficient of the LSP is calculated as shown in FIG.

[0102] Comparing the coefficients of the LSP after dimension conversion shown in Fig. 7 with the coefficients of the LSP of the same dimension calculated from the PCM data shown in Fig. 6, it can be seen that these coefficients are close to each other, and the difference between the coefficients becomes smaller as the dimension increases. Therefore, it can be said that the coefficients of the LSP obtained when the dimension is increased by the LSP dimension conversion process according to this embodiment are relatively accurate.

[0103] On the other hand, when an LSP of a predetermined dimension calculated using PCM data is lowered by one dimension through the LSP dimension conversion process according to this embodiment, each coefficient of the LSP is calculated as shown in FIG.

[0104] Comparing the coefficients of the LSP after dimension conversion shown in Fig. 8 with the coefficients of the LSP of the same dimension calculated from the PCM data shown in Fig. 6, it can be seen that these coefficients are close to each other, and that the difference between the coefficients becomes smaller as the dimension becomes higher. Therefore, it can be said that the coefficients of the LSP obtained when the dimension is lowered by the LSP dimension conversion process according to this embodiment are relatively accurate.

[0105] The above-mentioned encoder 10 and decoder 40 can each be configured by a computer device.

[0106] Fig. 9 is a configuration diagram of a computer device according to an embodiment. In this embodiment, the encoder 10 and the decoder 40 are configured as separate computer devices, but these computer devices may have the same configuration. Therefore, in the following description, for convenience, the computer device shown in Fig. 9 will be used to describe the computer device that configures the encoder 10 and the decoder 40.

[0107] The computer device 100 includes, for example, a central processing unit (CPU) 101, a main memory 102, a graphics processing unit (GPU) 103, a reader / writer 104, a communication interface (communication I / F) 105, an auxiliary storage device 106, an input / output interface (input / output I / F) 107, an output device 108, and an input device 109. The CPU 101, the main memory 102, the GPU 103, the reader / writer 104, the communication I / F 105, the auxiliary storage device 106, the input / output I / F 107, and the output device 108 are connected via a bus 110. The encoder 10 and the decoder 40 are each configured by appropriately selecting a part or all of the components described in the computer device 100.

[0108] In the computer device 100 constituting the encoder 10, a CPU 101 executes a processing program stored in an auxiliary storage device 106 to constitute, for example, a target data input unit 11, an LSP encoding processing unit 12, a target data encoding unit 13, and a compressed data output unit 14.

[0109] In the computer device 100 constituting the decoder 40, the CPU 101 executes a processing program stored in the auxiliary storage device 106 to constitute, for example, a compressed data input unit 41, an LSP decoding processing unit 42, a target data calculation unit 43, and a target data processing unit 44.

[0110] The main memory 102 is, for example, a RAM, a ROM, etc., and stores programs (processing programs, etc.) executed by the CPU 101 and various information. The auxiliary storage device 106 is, for example, a non-temporary storage device (non-volatile storage device) such as a Hard Disk Drive (HDD) or a Solid State Drive (SSD), and stores the programs executed by the CPU 101 and various information.

[0111] The GPU 103 is a processor that is suitable for executing a specific process, for example, a process that is performed in parallel. In this embodiment, the GPU 103 executes a predetermined process according to an instruction from the CPU 101.

[0112] The reader / writer 104 is detachably attached to the recording medium 111, and reads data from the recording medium 111 and writes data to the recording medium 111. The recording medium 111 may be, for example, a non-temporary recording medium (non-volatile recording medium) such as an SD memory card, an FD (floppy disk: registered trademark), a CD, a DVD, a BD (registered trademark), or a flash memory. In this embodiment, a processing program may be stored in the recording medium 111, and the reader / writer 104 may read and use the processing program. In the computer device 100 constituting the encoder 10, audio data to be processed may be stored in the recording medium 111, and the reader / writer 104 may read and use the processing program. In the computer device 100 constituting the encoder 10, the reader / writer 104 may store compressed data in the recording medium 111. In the computer device 100 constituting the decoder 40, the reader / writer 104 may read and use the compressed data from the recording medium 111.

[0113] The communication I / F 105 is connected to a network 60 , and transmits and receives data to and from other devices connected to the network 60 .

[0114] The input / output I / F 107 is connected to an input device 109 such as a mouse or a keyboard. In the computer device 100 constituting the encoder 10, the input / output I / F 107 accepts an operation input by a user of the encoder 10 using the input device 109. Also, in the computer device 100 constituting the decoder 40, the input / output I / F 107 accepts an operation input by a user of the decoder 40 using the input device 109.

[0115] The output device 108 is, for example, a display device such as a liquid crystal display or a speaker device, and outputs various types of information and sounds. The output device 108 is used, for example, in the decoder 40, for outputting sounds by the target data processing unit 44.

[0116] Next, a processing system according to a modified example will be described.

[0117] The processing system according to the modified example includes an LSP encoding processor 15 (see FIG. 10) instead of the LSP encoding processor 12 of the encoder 10 in the processing system 1 described above.

[0118] 10 is a functional block diagram of an LSP encoding unit of an encoder according to a modified example, in which the same components as those of the LSP encoding unit 12 shown in FIG.

[0119] The LSP encoding unit 15 is configured by further providing an LSP inverse conversion unit 31 (31-m, 31-m+1, etc.) for each section in the LSP encoding unit 12.

[0120] The LSP inverse transform unit 31 converts the LSP (fourth line spectrum pair) output by the adder unit 25 into an LPC (fourth linear prediction coefficient) of the same dimension, and outputs the converted LPC to the target data encoding unit 13. The target data encoding unit 13 encodes the section data string using the LPC (third linear prediction coefficient, an example of a third linear prediction coefficient corresponding coefficient) output from the LSP inverse transform unit 31. Here, when xm[t] is the first section, the LPC output from the LSP inverse transform unit 31-m is an example of a first linear prediction coefficient, a first linear prediction coefficient corresponding coefficient.

[0121] Here, the LPC output by the LSP inverse transform unit 31 has the same value as the LPC output from the LSP inverse transform unit 53 of the decoder 40. Therefore, the LPC used to encode the section data string on the encoder 10 side and the LPC used to decode the section data string on the decoder 40 side can be made to have the same value, and the decoder 40 side can appropriately restore data that is the same as the original voice data.

[0122] The present invention is not limited to the above-described embodiment and modified examples, and can be appropriately modified and implemented without departing from the spirit of the present invention.

[0123] For example, in the above embodiment, the LSP difference data is quantized, but the present invention is not limited to this, and the LSP difference data may not be quantized. In this case, the encoder 10 does not need to include the LSP quantizer 23 and the LSP inverse quantizer 24, and the decoder 40 does not need to include the LSP inverse quantizer 51.

[0124] In the above embodiment, the difference calculation unit 22, the addition unit 25, and the addition unit 52 use, for example, zero as a default value when the target section is the first section, but the default value is not limited to this, and may be, for example, an LPC in which all coefficients of the corresponding order are 0, converted into an LSP of the same order. In this case, the coefficients of the LSP are arranged at equal intervals between 0 and π.

[0125] In the above embodiment, an example has been shown in which the computer device 100 includes either the encoder 10 or the decoder 40. The present invention is not limited to this, and the computer device 100 may include both the encoder 10 and the decoder 40. [Explanation of symbols]

[0126] 1... processing system, 10... encoder, 11... target data input unit, 12... LSP encoding processing unit, 13... target data encoding unit, 14... compressed data output unit, 40... decoder, 41... compressed data input unit, 42... LSP decoding processing unit, 43... target data calculation unit, 44... target data processing unit, 60... network, 100... computer device, 101... CPU, 102... main memory, 108... output device, 111... recording medium

Claims

1. 1. A processing device utilizing line spectrum pairs, comprising: a first linear prediction coefficient conversion unit for converting a first-dimensional first line spectrum pair into a first-dimensional first linear prediction coefficient; a linear prediction coefficient dimension conversion unit that converts the first linear prediction coefficients of the first dimension into second linear prediction coefficients of a second dimension; a first line spectrum pair converter for converting the second linear prediction coefficients of the second dimension into a second line spectrum pair of the second dimension; A processing device having

2. When the second dimension is a higher dimension than the first dimension, the linear prediction coefficient dimension conversion unit converts the first linear prediction coefficient of the first dimension into the second linear prediction coefficient of the second dimension by adding coefficients that are zero as coefficients corresponding to a dimensional difference between the second dimension and the first dimension. The processing device of claim 1 .

3. When the second dimension is a dimension lower than the first dimension, the linear prediction coefficient dimension conversion unit converts the second dimension into the second linear prediction coefficient by repeating a prediction process of predicting a linear prediction coefficient of a dimension one dimension lower from the first linear prediction coefficient of the first dimension until the linear prediction coefficient becomes the second dimension. The processing device according to claim 1 or 2.

4. a determination unit that determines a third linear predictive coefficient of the second dimension that is applicable to a second section data sequence subsequent to a first section data sequence to be encoded using the first linear predictive coefficient of the first dimension or a linear predictive coefficient similar to the first linear predictive coefficient; a second line spectrum pair converter that converts the third linear prediction coefficients of the second dimension into a third line spectrum pair of the second dimension; a difference calculation unit that calculates difference correspondence data corresponding to a difference between the second spectrum pair and the third spectrum pair; a coding unit that codes the second interval data string using the third linear predictive coefficients to generate interval coded data; an output unit that outputs compressed data including the section encoded data and the difference corresponding data; The processing device of claim 3 .

5. a determination unit that determines a third linear predictive coefficient of the second dimension to be applied to a second section data string subsequent to a first section data string to be encoded using the first linear predictive coefficient of the first dimension; a second line spectrum pair converter that converts the third linear prediction coefficients of the second dimension into a third line spectrum pair of the second dimension; a difference calculation unit that calculates quantized difference data by quantizing a difference between the second spectral line pair and the third spectral line pair; an inverse quantization unit that generates inverse quantized difference data by inverse quantizing the quantized difference data; an adder that calculates a fourth spectral line pair by adding the inverse quantization difference data to the second spectral line pair; an inverse transform unit that transforms the fourth line spectrum pair into fourth linear prediction coefficients; a coding unit that codes the second interval data string using the fourth linear predictive coefficients to generate interval coded data; an output unit that outputs compressed data including the quantized difference data and the section encoded data; The processing device of claim 3 .

6. a compressed data receiving unit that receives compressed data including: (A) first section encoded data obtained by encoding a first section data string according to a first linear predictive coefficient corresponding to the first linear predictive coefficient of a first dimension; (B) second section encoded data obtained by encoding a second section data string subsequent to the first section data string according to a third linear predictive coefficient corresponding to a third linear predictive coefficient of the second dimension; (C) calculation data for calculating a first line spectrum pair of the first dimension corresponding to the first linear predictive coefficient corresponding coefficient; and (D) difference correspondence data corresponding to a difference between the second line spectrum pair of the second dimension obtained by converting the first linear predictive coefficient of the first dimension into a second linear predictive coefficient of the second dimension and converting the second linear predictive coefficient, and a third line spectrum pair corresponding to the third linear predictive coefficient corresponding coefficient. a first spectrum line pair calculation unit that calculates the first spectrum line pair based on the calculation data; a calculation unit that calculates third spectrum pair corresponding data corresponding to the third spectrum pair based on the second spectrum pair and the difference correspondence data; a second linear prediction coefficient conversion unit that converts the third line spectrum pair corresponding data into the second-dimensional coefficient corresponding to the third linear prediction coefficient; a restoration unit that performs a restoration process on the second section encoded data by using the third linear predictive coefficient corresponding coefficient of the second dimension; Equipped The processing device of claim 1 .

7. A processing method using a processing device that utilizes a line spectrum pair, comprising: Transforming a first pair of first line spectra in a first dimension into first linear prediction coefficients in the first dimension; Transforming the first linear prediction coefficients of the first dimension into second linear prediction coefficients of a second dimension; A processing method for converting the second linear prediction coefficients of the second dimension into a second line spectrum pair of the second dimension.

8. A processing program to be executed by a computer constituting a processing device using a line spectrum pair, The computer, a first linear prediction coefficient conversion unit for converting a first-dimensional first line spectrum pair into a first-dimensional first linear prediction coefficient; a linear prediction coefficient dimension conversion unit that converts the first linear prediction coefficients of the first dimension into second linear prediction coefficients of a second dimension; a first line spectrum pair converter that converts the second linear prediction coefficients of the second dimension into a second line spectrum pair of the second dimension; Processing program.

Images