Method and device for processing audio signals, audio decoder, and audio encoder

Abstract

To provide an improved method and device for removing discontinuities within filtered audio frames without creating any potential distortion in a filtered audio signal.SOLUTION: A method for processing an audio signal (100) removes discontinuities between a filtered past frame and a filtered present frame of the audio signal using linear predictive filtering (102, 110, 112).SELECTED DRAWING: Figure 5

Description

[Technical Field]

[0001] The present invention relates to the field of audio signals, and more particularly to a method for processing an audio signal comprising a plurality of audio frames, in which discontinuities between successive filtered audio frames are reduced or eliminated. [Background technology]

[0002] In the field of audio signal processing, audio signals may be filtered for various reasons, for example, long-term predictive filters may be used in audio signal encoders to completely attenuate or further suppress a set of harmonics in the audio signal.

[0003] An audio signal includes multiple audio frames, and the frames are filtered using a long-term prediction filter. When considering two consecutive frames of the audio signal, a past frame and a current frame, a linear filter H(z) with a set of parameters c is used to filter the audio signal. More specifically, the past frame is filtered with the filter H(z) using a first set of parameters c0 to produce a so-called filtered past frame. The current frame is filtered with the filter H(z) using a set of parameters c1 to produce a filtered current frame. FIG. 1 is a block diagram for processing consecutive frames of an audio signal according to a known method. An audio signal 100 including multiple audio frames is provided. The audio signal 100 is provided to a filter block 102, and a current frame n of the audio signal 100 is filtered. The filter block on the side of the audio signal 100 provides filter parameters c for the current frame of the audio signal. n, respectively. The filter block 102 filters a current frame n of the audio signal and outputs a filtered audio signal 104 comprising successively filtered frames. In FIG. 1, a filtered current frame n, a filtered past frame n-1, and a filtered penultimate frame n-2 are schematically represented. The filtered frames are schematically represented in FIG. 1 with respective gaps therebetween to schematically indicate discontinuities 106a, 106b that may be introduced by filtering between the filtered frames. The filter block 102 provides filtering of the frames of the audio signal using filter parameters c0 and c1 for the past frame n-1 and the current frame n, respectively. In general, the filter block 102 can be a linear filter H(z), and one example of such a linear filter H(z) is the long-term prediction filter (LPC) discussed above. H(z)=1-g z -T where the filter parameters are the gain 'g' and the pitch lag 'T'. In a more general form, the long-term prediction filter can be written as: H(z)=1-g A(z) z -T where A(z) is an FIR filter. A long-term prediction filter can be used to completely attenuate or further suppress a set of harmonics in an audio signal. However, when using this type of long-term prediction filter, and if the filter parameter c0 of the past frame is different from the filter parameter c1 of the current frame, there is a high probability of introducing a discontinuity 106a, 106b (see FIG. 1) between the filtered past frame n-1 and the filtered current frame n. This discontinuity can create artifacts, such as "clicks," in the filtered audio signal 104.

[0004] Therefore, in view of the above-mentioned problems with filtering successive frames resulting in discontinuities that can create unwanted artifacts, there is now a need for a technique for removing possible discontinuities. Several technical methods are known in the art that deal with the removal of discontinuities in filtered frames of an audio signal.

[0005] When the linear filter H(z) is an FIR filter, the current frame is filtered using the filter parameters c1 of the current frame to produce a filtered current frame. In addition, an initial portion of the current frame is filtered using the filter parameters c0 of the previous frame to produce a filtered frame portion, and then an overlap-add or cross-fade operation is performed through the filtered current frame and the initial portion of the filtered frame portion. FIG. 2 shows a block diagram of this type of conventional method for processing consecutive audio frames to remove discontinuities. Compared to FIG. 1, the filter block 102 includes an additional processing block 108 for performing the overlap-add or cross-fade operation. In the filtered audio signal 104, discontinuities between consecutive filtered frames n, n-1, and n-2 of FIG. 1 may be eliminated or reduced, as shown schematically in FIG. 2, which shows consecutive gap-free filtered frames n, n-1, and n-2.

[0006] In other prior art methods, the filter H(z) may be a filter with recursive elements, e.g., an IIR filter. In such cases, the method described above with reference to FIG. 2 is applied sample-by-sample. In a first step, processing begins with a first sample of the initial portion of current frame n being filtered using filter parameter c0 of previous frame n-1 to obtain a first filtered sample. The sample is filtered with filter parameter c1 of current frame n to produce a second filtered sample. Then, a convolution-add or crossfade operation is performed based on the first and second filtered samples obtained corresponding to the filtered sample of current frame n. Then, the next sample is processed, and the above steps are repeated until the last sample of the initial portion of current frame n is processed. The remaining samples of current frame n are processed using filter parameter c1 of current frame n.

[0007] Examples for the above-mentioned known methods for removing discontinuities from successive filtered frames are described, for example, in U.S. Pat. No. 5,012,517 in connection with a transform coder, in EP 0,732,687 in connection with a speech band extension, in U.S. Pat. No. 5,999,899 in connection with a transform audio coder, or in U.S. Pat. No. 7,353,168 in connection with a decoded speech postfilter.

[0008] While the above methods are effective for removing unwanted signal discontinuities, they operate on a specific portion of the current frame (the beginning) to be effective, so the length of the frame portion must be long enough. For example, for a frame length of 20 ms, the length of the frame portion or the beginning portion may be as long as 5 ms. In certain cases, this may be too long, especially in situations where the filter parameter c0 of the previous frame does not apply to the current frame, resulting in additional artifacts. One example is a harmonic audio signal with a rapidly changing pitch, and a long-term prediction filter is designed to reduce the amplitude of the harmonics. In this case, the pitch lag varies from one frame to the next. A long-term prediction filter with an estimated pitch in the current frame effectively reduces the amplitude of the harmonics in the current frame, but if it is used in another frame (e.g., the beginning of the next frame) where the pitch of the audio signal is different, it will not reduce the amplitude of the harmonics. Reducing the amplitude of non-harmonic-related components in the signal may even make things worse, introducing distortion into the signal. [Prior art documents] [Patent documents]

[0009] [Patent Document 1] U.S. Patent No. 5,012,517 [Patent Document 2] European Patent No. 0732687 [Patent Document 3] U.S. Patent No. 5,999,899 [Patent Document 4] U.S. Patent No. 7,353,168 Summary of the Invention [Problem to be solved by the invention]

[0010] It is a basic object of the present invention to provide an improved method for removing discontinuities in filtered audio frames without creating any potential distortions in the filtered audio signal.

[0011] This object is achieved by the methods and devices described in the independent claims.

[0012] The present invention provides a method for processing an audio signal, the method comprising the step of removing discontinuities between a filtered past frame and a filtered current frame of the audio signal using linear predictive filtering.

[0013] The linear prediction filter is TIFF2025186274000002.tif20159, M is the filter order, and a m are the filter coefficients (with a0 = 1). This type of filter is also known as Linear Predictive Coding (LPC).

[0014] According to an embodiment, the method comprises the steps of filtering a current frame of an audio signal and removing discontinuities by modifying a first portion of the filtered current frame with a signal obtained by linear predictive filtering of a predetermined signal, the initial state of the linear predictive filter being defined based on the last portion of a previous frame.

[0015] According to an embodiment, the initial state of the linear prediction filter is defined based on the last part of a previous unfiltered frame that has been filtered with a set of filter parameters for filtering the current frame.

[0016] According to an embodiment, the method comprises the step of estimating a linear prediction filter on a filtered or unfiltered audio signal.

[0017] According to an embodiment, estimating the linear prediction filter comprises estimating the filter based on a past or current frame of the audio signal or based on a filtered past frame of the audio signal using the Levinson-Durbin algorithm.

[0018] According to an embodiment, the linear prediction filter comprises a linear prediction filter of an audio codec.

[0019] According to an embodiment, removing discontinuities includes processing a first portion of the current filtered frame, the first portion of the current frame having a predetermined number of samples that is less than or equal to the total number of samples in the current frame, and processing the first portion of the current frame includes subtracting a first portion of a zero-input-response (ZIR) from the first portion of the current filtered frame.

[0020] According to an embodiment, the method includes filtering a current frame of an audio signal with a non-recursive filter, such as an FIR filter, to produce a filtered current frame.

[0021] According to an embodiment, the method comprises processing a current unfiltered frame of an audio signal sample by sample using a recursive filter, such as an IIR filter, and processing samples of an initial portion of the current frame comprises: filtering the samples with a recursive filter using filter parameters of the current frame to produce filtered samples; and subtracting the corresponding ZIR samples from the filtered samples to produce corresponding samples of the current filtered frame.

[0022] According to an embodiment, the filtering and subtracting steps are repeated until the last sample in the first part of the current frame has been processed, and the method further comprises filtering the remaining samples in the current frame with a recursive filter using the filter parameters of the current frame.

[0023] According to an embodiment, the method includes generating a ZIR, the generating ZIR comprising: filtering the M last samples of the unfiltered past frame with a filter and filter parameters used to filter the current frame to produce a first portion of a filtered signal, where M is the order of the linear prediction filter; subtracting the M last samples of the filtered past frame, which were filtered using the filter parameters of the past frame, from the first portion of the filtered signal to generate a second portion of the filtered signal; generating a ZIR for the linear prediction filter by filtering a frame of zero samples with an initial state equal to the linear prediction filter and a second portion of the filtered signal.

[0024] According to an embodiment, the method includes windowing the ZIR such that its amplitude rapidly decreases to zero.

[0025] The present invention is based on the inventor's discovery that the problems recognized in conventional methods for removing signal discontinuities, which result in the aforementioned additional and unnecessary distortion, are primarily due to the processing of a current frame or at least portions thereof based on filter parameters for past frames. The method of the present invention avoids this problem. That is, the method of the present invention does not filter portions of a current frame using filter parameters from past frames, thereby avoiding the aforementioned problems. According to an embodiment, a linear predictive filter (LPC) is used to remove the discontinuities. The LPC filter is estimated on the audio signal and therefore is a good model of the spectral shape of the audio signal, so that the LPC filter masks the discontinuities. In an embodiment, the LPC filter can be estimated based on an unfiltered audio signal or based on an audio signal filtered by the aforementioned linear filter H(z). According to an embodiment, the LPC filter can be estimated using the audio signal, e.g., the current frame and / or past frames, and the Levinson-Durbin algorithm. It can also be calculated based only on the filtered past frame signal using the Levinson-Durbin algorithm.

[0026] Furthermore, in other embodiments, an audio codec for processing an audio signal may use a linear filter H(z) and, for example, a quantized or unquantized LPC filter to shape the quantization noise in a transform-based audio codec. In such embodiments, this existing LPC filter may be used directly to smooth discontinuities without the additional complexity required to estimate a new LPC filter.

[0027] In the following, embodiments of the present invention will be described with reference to the accompanying drawings. [Brief explanation of the drawings]

[0028] [Figure 1] FIG. 1 shows a block diagram for processing successive frames of an audio signal according to a conventional method. [Figure 2] FIG. 2 shows a block diagram of another conventional method for processing successive audio frames to remove discontinuities. [Figure 3] FIG. 3 shows a simplified block diagram of a system for transmitting an audio signal implementing the inventive method for removing discontinuities between successive frames of an audio signal at the encoder side and / or decoder side. [Figure 4] FIG. 4 shows a flow diagram illustrating the inventive method for removing discontinuities between successive frames of an audio signal according to an embodiment. [Figure 5] FIG. 5 shows a schematic block diagram for processing a current audio frame according to an embodiment of the invention, which avoids unwanted distortions in the output signal despite the removal of discontinuities. [Figure 6] FIG. 6 shows a flow diagram illustrating the functioning of the blocks of FIG. 5 for generating a ZIR. [Figure 7] FIG. 7 shows a flow diagram illustrating the functioning of the blocks in FIG. 5 for processing the first part of the current filtered frame when the filter block includes a recursive filter such as an IIR filter. [Figure 8] FIG. 8 shows a flow diagram illustrating the functioning of the blocks in FIG. 5 for processing the first part of the current filtered frame when the filter block includes a non-recursive filter such as an FIR filter. DETAILED DESCRIPTION OF THE INVENTION

[0029] In the following, embodiments of the inventive method will be described in more detail, and it should be noted that in the accompanying drawings, elements having the same or similar functions are denoted by the same reference numerals.

[0030] FIG. 3 shows a simplified block diagram of a system for transmitting an audio signal implementing the inventive method on the encoder side and / or decoder side. The system of FIG. 3 includes an encoder 200 receiving an audio signal 204 at an input device 202. The encoder includes a coding processor 206 receiving the audio signal 204 and generating an encoded audio signal provided at an output device 208 of the encoder. The coding processor may be programmed or configured to implement the inventive method for processing successive audio frames of the received audio signal to avoid discontinuities. In other embodiments, the encoder need not be part of a transmission system, but may be a stand-alone device generating the encoded audio signal or may be part of an audio signal transmitter. According to an embodiment, the encoder 200 may include an antenna 210, as indicated at 212, for allowing wireless communication of the audio signal. In other embodiments, the encoder 200 may output the encoded audio signal provided at an output device 208, for example, using a wired connection, as indicated at 214.

[0031] 3 further includes a decoder 250 having an input device 252 that receives, for example, via wire 214 or antenna 254, the encoded audio signal processed by encoder 250. Encoder 250 includes a decoding processor 256 that operates on the encoded signal and provides a decoded audio signal 258 at output device 260. Decoding processor 256 may be implemented to operate according to the method of the invention on successive frames that are filtered in such a way that discontinuities are avoided. In other embodiments, the decoder need not be part of the transmission system, but rather may be a stand-alone device for decoding the encoded audio signal, or may be part of an audio signal receiver.

[0032] In the following, an embodiment of the inventive method that may be implemented in at least one of the encoding processor 206 and the decoding processor 256 will be described in more detail. Figure 4 shows a flow diagram for processing a current frame of an audio signal according to an embodiment of the inventive method. Processing of the current frame will be described, and it is assumed that past frames have already been processed using the same technique described below. According to the invention, in step S100, a current frame of an audio signal is received. The current frame is filtered in step S102, for example, in the manner as described above with respect to Figures 1 and 2 (see filter block 102). According to the inventive method, discontinuities between the filtered past frame n-1 and the filtered current frame n (see Figure 1 or Figure 2) are removed using linear predictive filtering, as shown in step S104. According to an embodiment, the linear predictive filter TIFF2025186274000003.tif22111, M is the filter order, and a mare filter coefficients (with a0 = 1). This type of filter is also known as Linear Predictive Coding (LPC). According to an embodiment, the filtered current frame is processed by applying linear predictive filtering to at least a portion of the filtered current frame. Discontinuities may be removed by modifying an initial portion of the filtered current frame with a signal obtained by linear predictive filtering of a predetermined signal, where the initial state of the linear predictive coding filter is defined based on the final portion of the previous frame. The initial state of the linear predictive coding filter may be defined based on the final portion of the previous frame filtered with the set of filter parameters for the current frame. Advantageously, the methods of the present invention do not need to filter the current frame of the audio signal with the filter coefficients used for the previous frame, thereby avoiding problems that arise due to an inappropriate combination of filter parameters for the current and previous frames, as experienced in the prior art method described above with reference to FIG. 2.

[0033] FIG. 5 shows a schematic block diagram for processing a current audio frame of an audio signal according to an embodiment of the present invention, which removes discontinuities while avoiding unwanted distortion in the output signal. In FIG. 5, the same reference numerals as in FIGS. 1 and 2 are used. A current frame n of an audio signal 100 is received, and each frame of the audio signal 100 has a plurality of samples. The current frame n of the audio signal 100 is processed by a filter block 102. Compared with the prior art method of FIGS. 1 and 2, according to the embodiment described with reference to FIG. 5, the filtered current frame is further processed based on ZIR samples, as indicated generally by block 110. According to this embodiment, based on the past frame n-1 and based on an LPC filter, ZIR samples are produced, as indicated generally by block 112.

[0034] The functions of processing blocks 110 and 112 will now be described in more detail. FIG. 6 shows a flow diagram illustrating the function of processing block 112 for generating ZIR samples. As previously described, frames of audio signal 100 are filtered by a linear filter H(z) using filter parameters c selected or determined for each frame. Filter H(z) can be a recursive filter, e.g., an IIR filter, or a non-recursive filter, e.g., an FIR filter. In processing block 112, a quantized or unquantized LPC filter is used. The LPC filter has order M and can be estimated on a filtered or unfiltered audio signal or an LPC filter used in an audio codec. In step S200, the M (M = LPC filter order) last samples of past frame n-1 are filtered by filter H(z) using filter parameters or coefficients c1 for the current frame n. Step S200 thereby produces a first portion of the filtered signal. In step S202, the M last samples of the filtered past frame n-1 (the M last samples of the past frame filtered using the filter parameters or coefficients c0 of the past frame n-1) are subtracted from the first portion of the filtered signal provided by step S200, thereby producing a second portion of the filtered signal. In step S204, an LPC filter having order M is applied; more specifically, a zero-input response (ZIR) of the LPC filter is generated in step S204 by filtering a frame of zero samples. Here, the initial state of the filter is equal to the second portion of the filtered signal, thereby producing the ZIR. According to an embodiment, the ZIR may be windowed such that its amplitude rapidly decreases to zero.

[0035] As described above with respect to FIG. 5, ZIR is applied in processing block 110, and its function is described with respect to the flow diagram of FIG. 7 for the case where a recursive filter, such as an IIR filter, is used as the linear filter H(z). According to the embodiment described with respect to FIG. 5, in order to avoid undesired distortions and remove discontinuities between the current frame and past frames, the filtering step for current frame n includes processing (filtering) current frame n sample by sample. Here, the first portion of samples is processed according to the method of the present invention. More specifically, the first M samples of current frame n are processed. In the first step S300, a variable m is set to 0. In the next step S302, sample m of current frame n is filtered using filter H(z) and filter coefficients or parameters c1 for current frame n. Thus, unlike conventional methods, according to the method of the present invention, the current frame is not filtered using coefficients from past frames. As a result, despite the fact that only coefficients from the current frame are used to remove discontinuities, the undesired distortions present in conventional methods are avoided. Step S302 obtains filtered sample m, and in step S304, the ZIR sample corresponding to sample m is subtracted from filtered sample m to obtain the corresponding sample of filtered current frame n. In step S306, it is determined whether the last sample M of the first portion of current frame n has been processed. If not all M samples of the first portion have been processed, variable m is incremented, and method steps S302 to S306 are repeated for the next sample of current frame n. Once all M samples of the first portion have been processed, in step S308, the remaining samples of current frame n are filtered using the filter parameters c1 of the current frame, thereby providing filtered current frame n, which is processed according to the method of the invention, avoiding undesired artifacts in removing discontinuities between successive frames.

[0036] According to another embodiment, the linear filter H(z) is a non-recursive filter, such as an FIR filter, and ZIR is applied in processing block 110, as described above with respect to FIG. 5. The functionality of this embodiment is described with respect to the flow diagram of FIG. 8. In step S400, the current frame n is filtered with the filter H(z) using the filter coefficients or parameters c1 for the current frame. Thus, in accordance with an embodiment of the present invention, in addition to conventional methods, the current frame is not filtered using coefficients from past frames, thereby avoiding undesired distortions present in conventional methods despite the fact that only coefficients from the current frame are used to remove discontinuities. In step S402, the initial portion of the ZIR is subtracted from the corresponding initial portion of the filtered current frame, thereby providing a filtered current frame n having an initial portion filtered / processed according to the method of the present invention, and leaving only the portion filtered using the filter coefficients or parameters c1 for the current frame, thereby avoiding undesired distortions due to the removal of discontinuities between successive frames.

[0037] The method of the invention can be applied in situations such as those described above when the audio signal is filtered. According to an embodiment, the method of the invention can also be applied on the decoder side, for example when using an audio codec post-filter to reduce the level of coding noise among signal harmonics. For processing audio frames on the decoder side, according to an embodiment, the post-filter can be of the form: H(z)=(1-B(z)) / (1-A(z) Z -T ) where B(z) and A(z) are two FIR filters, the H(z) filter parameters are the coefficients of the FIR filters B(z) and A(z), and T denotes the pitch lag. In this type of scenario, if, for example, the past filter frame parameter c0 differs from the current frame filter parameter c1, the filters may introduce a discontinuity between the two filtered frames, and this type of discontinuity may create artifacts in the filtered audio signal 104, such as a "click." This discontinuity is removed by processing the current filtered frame, as described in detail above.

[0038] Although some aspects of the described concepts are described in the context of an apparatus, it will be apparent that these aspects also represent descriptions of corresponding methods, where blocks or devices correspond to method steps or features of method steps. Analogously, aspects described in the context of method steps represent descriptions of corresponding blocks, items, or features of apparatus.

[0039] Depending on specific implementation requirements, embodiments of the present invention may be implemented in hardware or software. The implementation may be implemented using a digital storage medium, such as a floppy disk, DVD, Blu-ray disk, CD, ROM, PROM, EPROM, EEPROM, or FLASH memory, having electronically readable control signals stored thereon, which cooperate (or can cooperate) with a programmable computer system so that the respective methods are performed. Thus, the digital storage medium may be computer-readable.

[0040] Some embodiments according to the present invention include a data carrier having an electronically readable signal that can cooperate with a programmable computer system to perform one of the methods described herein.

[0041] Typically, embodiments of the present invention are implemented as a computer program product having program code which, when the computer program product runs on a computer, is activated to perform one of the methods. The program code may, for example, be stored on a machine-readable carrier.

[0042] Other embodiments comprise the computer program for performing one of the methods described herein, stored on a machine readable carrier.

[0043] In other words, therefore, an embodiment of the inventive method is a computer program comprising a program code for performing one of the methods described herein, when the computer program runs on a computer.

[0044] A further embodiment of the inventive method is, therefore, a data carrier (or digital storage medium, or computer readable medium) comprising, recorded on it, a computer program for performing one of the methods described herein.

[0045] A further embodiment of the inventive method is, therefore, a data stream or sequence of signals representing the computer program for performing one of the methods described herein, for example a data stream or sequence of signals adapted to be transmitted via a data communications connection, for example the Internet.

[0046] A further embodiment comprises a processing means, for example a computer, or a programmable logic circuit, configured to or adapted to perform one of the methods described herein.

[0047] A further embodiment comprises a computer having installed thereon the computer program for performing one of the methods described herein.

[0048] In some embodiments, programmable logic circuitry (e.g., a Field Programmable Gate Array (FPGA)) may be used to perform some or all of the functions described herein. In some embodiments, the Field Programmable Gate Array may cooperate with a microprocessor to perform one of the methods described herein. In general, the methods are preferably performed by some hardware device.

[0049] The above-described embodiments merely represent examples of the principles of the present invention. It is understood that modifications and variations of the devices described herein will be apparent to others skilled in the art. It is the intention, therefore, to be limited only by the scope of the appended claims and not by the detailed description of the present invention as set forth in the description and explanation thereof.

Claims

1. A method for processing an audio signal (100), said method comprising:

1. A method comprising the steps of: removing (S102, S104, S300-S308, S400-S402) discontinuities (106a, 106b) between filtered past frames and filtered current frames of the audio signal using linear predictive filtering.

2. filtering the current frame of the audio signal; and removing the discontinuity by modifying a first portion of the filtered current frame with a signal obtained by linear predictive filtering a predetermined signal, the initial state of the linear predictive filter being defined based on a last portion of the previous frame; The method of claim 1 , comprising:

3. 3. The method of claim 2, wherein the initial state of the linear prediction filter is defined based on a last portion of the unfiltered previous frame that was filtered using the set of filter parameters for filtering the current frame.

4. The method of any of claims 1 to 3, further comprising the step of estimating the linear prediction filter for the filtered or unfiltered audio signal (100).

5. 5. The method of claim 4, wherein estimating the linear predictive filter comprises estimating the linear predictive filter based on the past and / or current frames of the audio signal (100) or based on filtered past frames of the audio signal (100) using a Levinson-Durbin algorithm.

6. The method of claim 1 , wherein the linear prediction filter comprises a linear prediction filter of an audio codec.

7. 7. The method of claim 1, wherein the step of removing discontinuities includes processing the initial portion of the filtered current frame, the initial portion of the current frame having a predetermined number of samples that is less than or equal to the total number of samples in the current frame, and wherein the step of processing the initial portion of the current frame includes the step of subtracting (S304, S402) an initial portion of a zero-input response (ZIR) from the initial portion of the filtered current frame.

8. 8. The method of claim 7, comprising the step of filtering (S400) the current frame of the audio signal with a non-recursive filter, such as an FIR filter, to produce the filtered current frame.

9. processing the current unfiltered frame of the audio signal sample by sample using a recursive filter, such as an IIR filter, wherein processing the samples of the initial portion of the current frame comprises: filtering the samples with the recursive filter using the filter parameters of the current frame to produce filtered samples (S302); (S304) subtracting corresponding ZIR samples from the filtered samples to produce the corresponding samples of the current frame that have been filtered; The method of claim 7, comprising:

10. 10. The method of claim 9, wherein the filtering (S302) and subtracting (S304) steps are repeated until the last sample in the initial portion of the current frame has been processed, and the method further comprises the step of filtering (S306) the remaining samples in the current frame with the recursive filter using the filter parameters of the current frame.

11. generating the ZIR; The step of generating the ZIR comprises: filtering (S200) the M last samples of the unfiltered past frame using the filter and the filter parameters used to filter the current frame to produce a first portion of a filtered signal, where M is the order of the linear prediction filter; subtracting (S202) the M last samples of the filtered past frame, which have been filtered using the filter parameters of the past frame, from the first portion of the filtered signal to generate a second portion of the filtered signal; generating a ZIR of the linear prediction filter by filtering a frame of zero samples using an initial state equal to the linear prediction filter and the second portion of the filtered signal (S204); 11. The method of any one of claims 7 to 10, comprising:

12. 12. The method of claim 11, comprising windowing the ZIR such that its amplitude rapidly decreases to zero.

13. A non-transitory computer program product comprising a computer readable medium storing instructions that, when executed on a computer, perform the method of any of claims 1 to 12.

14. An apparatus for processing an audio signal (100), said apparatus comprising: An apparatus comprising: a processor (102, 110, 112) for removing discontinuities between filtered past frames and filtered current frames of the audio signal using linear predictive filtering.

15. Apparatus for processing an audio signal (100), said apparatus configured to operate according to the method of one of claims 1 to 12.

16. An audio decoder (250) comprising the device of claim 14 or claim 15.

17. An audio encoder (200) comprising the apparatus of claim 14 or claim 15.

Images