JP6652469B2 - Decoding device, decoding method, and program - Google Patents

Description

  The present invention relates to a technique for decoding a time-series signal such as a sound signal.

  Conventionally, as a countermeasure when a correct decoded sound signal cannot be obtained due to loss of information to be input to the decoding device due to loss of a packet on a transmission path between the encoding and decoding of the sound signal and the like, At the encoding stage, sound signal classification information such as signal energy, phase, and pitch period is supplementarily added, and based on auxiliary information input to the decoding device before loss of information to be input to the decoding device occurs. Then, by obtaining a signal of the same kind as the classification of the sound signal indicated by the auxiliary information, a decoded sound signal in a portion where information is missing is generated by interpolation (for example, see Non-Patent Document 1).

J. Lecomte, T. Vaillancourt, S. Bruhn, H. Sung, K. Peng, K. Kikuiri, B. Wang, S. Subasingha, and J. Faure, “Packet-loss concealment technology advances in EVS,” in Proc . ICASSP 2015, pp. 5708-5712, 2015.

  However, in the above technique, although the auditory discomfort at the time of information loss is reduced, an additional amount of information is required in addition to the information necessary for decoding a normal signal. Further, in the above technique, since a sound signal included in one packet has a length of about 20 msec, an additional principle delay for using information after a missing signal is necessary. there were.

  An object of the present invention is to provide a decoding device, a decoding method, and a program that can interpolate missing information more audibly than conventional techniques without requiring additional information transmission and without increasing the principle delay of decoding. I do.

A decoding device according to one aspect of the present invention is a decoding device that obtains a decoded sound signal for each frame. For a frame in which a sound signal code is missing, a sample of a decoded sound signal of a frame preceding the frame is temporally extracted. A signal candidate having high temporal continuity with the previous frame is selected from the reversely arranged signal and the signal obtained by inverting the polarity of the signal, and the extended decoded sound signal which is a selected signal candidate and the previous frame are selected. And a signal generated on the basis of the signal obtained by linear prediction synthesis from the input signal and a decoded signal of the frame.

A decoding device according to one aspect of the present invention is a decoding device that obtains a decoded sound signal for each frame. For a frame in which a sound signal code is missing, a sample of a decoded sound signal of a frame preceding the frame is temporally extracted. Interpolation signal generation using a signal generated based on an extended decoded sound signal that is a signal arranged in reverse or a signal whose polarity is inverted and a signal that is linearly predicted and synthesized from the previous frame as a decoded sound signal of a frame Including parts.
A decoding device according to one aspect of the present invention is a decoding device that obtains a decoded sound signal for each frame. For a frame in which a sound signal code is missing, a sample of a decoded sound signal of a frame preceding the frame is temporally extracted. Among the reversely arranged signals, the signal whose polarity is inverted, the decoded sound signal of the previous frame, and the signal whose polarity of the decoded sound signal of the previous frame is inverted, the similarity to the signal obtained by linear predictive synthesis from the previous frame. A candidate for a high signal is selected, and an extended decoded sound signal that is a candidate for the selected signal, or a signal generated based on the extended decoded sound signal and a signal that is linearly predicted and synthesized from the previous frame, and a decoded sound signal of the frame And an interpolation signal generation unit for performing the interpolation.
A decoding device according to one aspect of the present invention is a decoding device that obtains a decoded sound signal for each frame. For a frame in which a sound signal code is missing, a sample of a decoded sound signal of a frame preceding the frame is temporally extracted. A signal having a high similarity to a signal obtained by linear prediction synthesis from the previous frame from a signal arranged in reverse or a signal in which the polarity of the signal is inverted is selected, and the selected decoded signal is an extended decoded sound signal, or an extended decoded sound signal. A signal generated based on a signal subjected to linear prediction synthesis from the previous frame, and an interpolated signal generation unit that uses the signal as a decoded sound signal of the frame.
A decoding device according to one aspect of the present invention is a decoding device that obtains a decoded sound signal for each frame. For a frame in which a sound signal code is missing, the decoded sound signal of the frame preceding the frame has the same power spectrum as the decoded sound signal. From among a plurality of signal candidates, a signal candidate having high temporal continuity with the previous frame is selected, and an extended decoded sound signal that is a candidate of the selected signal, and a signal obtained by linear prediction synthesis from the previous frame, And an interpolation signal generation unit that uses the signal generated based on the above as a decoded sound signal of the frame.
A decoding device according to one aspect of the present invention is a decoding device that obtains a decoded sound signal for each frame. For a frame in which a sound signal code is missing, the decoded sound signal of the frame preceding the frame has the same power spectrum as the decoded sound signal. From among a plurality of signal candidates, a signal candidate having a high similarity to the signal obtained by linear prediction synthesis from the previous frame is selected, and the extended decoded sound signal or the extended decoded sound signal that is the selected signal candidate is selected. An interpolated signal generation unit that uses a signal generated based on a signal obtained by performing linear prediction synthesis from the frame as a decoded sound signal of the frame.

  Missing information can be audibly interpolated better than in the prior art, without the need to transmit additional information and without increasing the decoding delay.

FIG. 3 is a block diagram illustrating an example of an encoding device assumed by a decoding device. 5 is a flowchart illustrating an example of an encoding method assumed by a decoding device. The figure which shows the example of a packet. FIG. 2 is a block diagram illustrating an example of a decoding device according to the first embodiment. 9 is a flowchart illustrating an example of a decoding method according to the first to third embodiments. FIG. 13 is a block diagram illustrating an example of a decoding device according to a second embodiment. FIG. 13 is a block diagram illustrating an example of a decoding device according to a third embodiment. The figure for demonstrating a technical background.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings.

[Encoding device assumed in first to third embodiments]
FIG. 1 shows an example of an encoding device assumed in a first embodiment to be described later and in a third embodiment. The assumed coding apparatus includes, for example, a linear prediction analysis unit 11, a signal coding unit 12, and a packetization unit 13, as shown in FIG.

  The encoding method assumed in the first to third embodiments is realized, for example, by each unit of the encoding device performing the processing of FIG. 2 and steps E1 to E3 described below.

  Hereinafter, each unit of the encoding device shown in FIG. 1 will be described.

<Linear prediction analysis unit 11>
A time-domain sound signal is input to the linear prediction analysis unit 11. The sound signal is, for example, an audio signal or an acoustic signal.

The linear prediction analysis unit 11 performs linear prediction coefficients α ₁ , α ₂ ,..., Α on the basis of a time-domain sound signal input in a frame unit of a predetermined time length corresponding to the time length of the sound signal included in the packet. Generate _p (step E1). Further, the linear prediction analysis unit 11 encodes the generated linear prediction coefficients α ₁ , α ₂ ,..., Α _p to obtain a linear prediction coefficient code. Examples of the linear prediction coefficient code, the linear predictive coefficients α _1, α _2, ..., a LSP code is a code corresponding to the column of the quantized value of the LSP (Line Spectrum Pairs) parameter sequence corresponding to alpha _p. p is an integer of 2 or more.

  The linear prediction analysis unit 11 outputs the obtained linear prediction coefficient code to the packetization unit 13.

Also, the linear prediction analyzer 11 is a linear prediction coefficient corresponding to the linear prediction coefficient code obtained quantized linear prediction coefficient _{^ α 1, ^ α 2,} ..., obtaining ^ alpha _p. It should be noted that the notation “^” with “•” as an arbitrary character means that “^” is added above “•”.

Linear prediction analysis unit 11, resulting quantized linear prediction coefficient _{^ α 1, ^ α 2,} ..., a ^ alpha _p, and outputs the signal encoding unit 12.

  The linear prediction analysis unit 11 performs a linear prediction analysis process, for example, by obtaining an autocorrelation with respect to a sound signal input in a frame unit and performing a Levinson-Durbin algorithm using the obtained autocorrelation. Is used. The acquisition of the linear prediction coefficient code by the linear prediction analysis unit 11 is performed by, for example, a conventional encoding technique. Conventional coding techniques include, for example, an encoding technique in which a code corresponding to the linear prediction coefficient itself is a linear prediction coefficient code, and a code corresponding to the LSP parameter obtained by converting the linear prediction coefficient into an LSP parameter. There are coding techniques for coding, and coding techniques for converting a linear prediction coefficient into a PARCOR coefficient and setting a code corresponding to the PARCOR coefficient as a linear prediction coefficient code.

<Signal encoding unit 12>
The signal encoding unit 12 receives a sound signal in the time domain and the quantized linear prediction coefficients ^ α ₁ , ^ α ₂ , ..., ^ α _p output from the linear prediction analysis unit 11.

The signal encoding unit 12 uses the values of the quantized linear prediction coefficients ^ α ₁ , ^ α ₂ ,..., ^ Α _p as in the encoding device of Non-Patent Document 1, for example, to input a sound input in frame units. Linear prediction of the signal is performed to obtain a prediction residual signal that is a linear prediction residual, and the obtained prediction residual signal is encoded to obtain a residual signal code (step E2). The signal encoding unit 12 outputs the obtained residual signal code to the packetizing unit 13.

<Packetizing unit 13>
The packetizer 13 receives the linear prediction coefficient code output from the linear prediction analyzer 11 and the residual signal code output from the signal encoder 12. Here, the input linear prediction coefficient code and residual signal code are codes representing the sound signal input to the encoding device, and thus the combination of the linear prediction coefficient code and the residual signal code is used as the sound signal code. Shall be called.

  For example, as shown in FIG. 3, the packetizing unit 13 includes a header indicating a data length and a frame number, a sound signal code, and an error such as a CRC code for detecting whether an error has occurred in the entire packet. One packet including the detection code is formed, and this packet is output to the decoding device (step E3). In this example, the sound signal code is a code including a linear prediction coefficient code and a residual signal code.

  If the data length of the audio signal code is fixed, it is not necessary to include information indicating the data length in the header, but the data length of the audio signal code differs depending on the packet when the signal encoding unit 12 performs variable-length encoding. Therefore, it is necessary to include information indicating the data length in the header.

[Decoding device and method according to first embodiment]
When a packet that is information to be input to the decoding device is lost due to packet loss or the like, the decoding device may use the packet (hereinafter, referred to as “missing packet”) in order to reduce auditory discomfort. ) Is generated from the information included in the packet that has not lost the decoded sound signal of the frame (hereinafter, the term “missing frame” may be used in some cases), that is, generated by interpolation. At this time, in order to perform the interpolation of the missing frame without increasing the principle delay of the decoding, the decoding device needs to interpolate from the information of the packet that is earlier in time than the missing packet of the missing frame. Here, the decoding device according to the first embodiment uses a signal predicted from the decoded linear prediction coefficient of the previous frame of the missing frame (hereinafter, sometimes referred to as a “predicted signal”) for interpolation of the missing frame. Used. That is, the decoding device of the first embodiment uses the prediction signal as the decoded sound signal of the missing frame from the decoded linear prediction coefficient of the previous frame, so that the continuity between the decoded sound signal of the missing frame and the decoded sound signal of the previous frame is obtained. To secure.

  However, the predicted signal attenuates with time, and if interpolation is performed using only the predicted signal, the decoded sound signal of the missing frame and the decoded sound signal of the next frame may be discontinuous. Therefore, the decoding device of the first embodiment uses a signal obtained by weighted sum of a predicted signal of a missing frame and a decoded sound signal of a previous frame for interpolation of the missing frame. That is, the decoding device of the first embodiment obtains, as a decoded sound signal of the missing frame, a signal obtained by weighting the sum of the predicted signal of the missing frame and the decoded sound signal of the previous frame. At this time, the decoding device of the first embodiment may use a decoded sound signal of the previous frame that has been subjected to time reversal or polarity inversion instead of the decoded sound signal of the previous frame. The decoding device according to the first embodiment uses the presence / absence of time reversal and the presence / absence of polarity reversal adaptively according to the decoded sound signal of the previous frame, thereby performing more continuous interpolation with the decoded sound signal of the previous frame. May go. In short, the decoding device of the first embodiment sets the signal obtained by the weighted sum of the predicted signal of the missing frame and the signal having the same power spectrum as the decoded sound signal of the previous frame as the decoded sound signal of the missing frame. Is also good.

  Further, the decoding apparatus of the first embodiment performs decoding of the next frame on the sound signal obtained by decoding the next frame in order to eliminate discontinuity between the prediction signal of the missing frame and the decoded sound signal of the next frame. By making the weighted sum of the obtained sound signal and the decoded sound signal of the missing frame the decoded sound signal of the next frame, a decoded sound signal continuous with the missing frame in the next frame may be obtained.

  FIG. 4 shows a configuration example of the decoding device of the first embodiment. As shown in FIG. 4, the decoding device according to the first embodiment includes, for example, a non-packetizing unit 21, a linear prediction coefficient decoding unit 22, a signal decoding unit 23, and an interpolation signal generation unit 24. The interpolation signal generation unit 24 includes, for example, a linear prediction coefficient storage unit 241, a decoded sound signal storage unit 242, a linear prediction unit 243, a signal extension selection unit 244, and an interpolation signal combination unit 245.

  The decoding method is realized, for example, by each unit of the decoding device that obtains a decoded sound signal for each frame, performing the processing of FIG. 5 and steps D1 to D6 described below.

  Hereinafter, each unit of the decoding device in FIG. 4 will be described.

<Non-packetizing unit 21>
The packet output from the encoding device is input to the non-packetizing unit 21.

  The depacketizing unit 21 detects whether an error has occurred in the packet based on an error detection code such as a CRC code in the packet. The non-packetizing unit 21 also determines whether or not a series of packet numbers are missing based on the frame number in the header of the packet, that is, among the sound signal codes of a plurality of frames that should exist continuously. It also detects whether a packet of a certain frame is missing due to packet loss or the like. Then, from those detection results, the non-packeting unit 21 determines that the sound signal code of the frame corresponding to the packet is not missing when the packet is not missing and there is no error in the packet, If the packet is missing or if there is an error in the packet, it is determined that the sound signal code of the frame corresponding to the packet is missing, and whether the sound signal code of the frame corresponding to the packet is missing Missing determination information, which is information indicating whether or not there is, is generated (step D1).

  The non-packeting unit 21 outputs the missing information to the linear prediction coefficient storage unit 241, the decoded sound signal storage unit 242, the linear prediction unit 243, the signal extension selection unit 244, and the interpolation signal combining unit 245. .

  The non-packeting unit 21 extracts a sound signal code from the packet based on the information indicating the data length in the header from the frame in which the sound signal code is not missing. , And at least the residual signal code of the sound signal codes to the signal decoding unit 23.

<Linear prediction coefficient decoding unit 22>
The linear prediction coefficient code output from the non-packetizing unit 21 for the frame in which the sound signal code is not missing is input to the linear prediction coefficient decoding unit 22.

The linear prediction coefficient decoding unit 22 decodes the input linear prediction coefficient code for each frame by using, for example, a conventional decoding technique for a frame in which the sound signal code is not missing, and decodes the decoded linear prediction coefficient ^ α ₁ , ^ α _2, ..., ^ to obtain alpha _p, resulting decoded linear prediction coefficient _{^ α 1, ^ α 2,} ..., and outputs a ^ alpha _p to the signal decoding unit 23 and the linear prediction coefficient storage unit 241 (step D2) .

  Here, the conventional decoding technique is, for example, when the linear prediction coefficient code is a code corresponding to the quantized linear prediction coefficient, the linear prediction coefficient code is the same as the quantized linear prediction coefficient. Technology to obtain decoded linear prediction coefficients, technology to obtain the same decoded LSP parameters as quantized LSP parameters by decoding the linear prediction coefficient codes when the linear prediction coefficient codes are codes corresponding to the quantized LSP parameters It is. Further, the linear prediction coefficient and the LSP parameter are mutually convertible, and the conversion between the decoded linear prediction coefficient and the decoded LSP parameter is performed according to the input linear prediction coefficient code and information necessary for processing in the subsequent stage. It is well known that the processing may be performed. As described above, what includes the decoding processing of the linear prediction coefficient code and the conversion processing performed as necessary is “decoding by the conventional decoding technique”.

  Note that the linear prediction coefficient decoding unit 22 does nothing with respect to the frame in which the sound signal code is missing.

<Signal decoding unit 23>
The signal decoding unit 23 outputs the residual signal code output from the non-packetizing unit 21 and the decoded linear prediction coefficient ^ α ₁ output from the linear prediction coefficient decoding unit 22 for the frame in which the sound signal code is not missing. ^ α ₂ , ..., ^ α _p are input.

The signal decoding unit 23 obtains a residual signal corresponding to the residual signal code for each frame, for example, as in the decoding device of Non-Patent Document 1, for a frame in which the sound signal code is not missing, and performs decoding linear prediction. By performing linear prediction synthesis using the coefficients ^ α ₁ , ^ α ₂ , ..., ^ α _p , the residual signal and the decoded sound signal up to one sample before, the decoded sound signal ^ x (0), ^ x ( 1),..., ^ X (N-1), and obtains the decoded sound signals ^ x (0), ^ x (1), ..., ^ x (N-1). The signal is output to the interpolation signal combining unit 245 (step D3). N is a predetermined positive integer.

  Note that the signal decoding unit 23 does nothing with respect to the frame in which the sound signal code is missing.

<Linear prediction coefficient storage unit 241>
The linear prediction coefficient storage unit 241 receives the loss determination information output by the depacketizing unit 21 and the decoded linear prediction coefficients ^ α ₁ , ^ α ₂ ,..., ^ Α output by the linear prediction coefficient decoding unit 22. Is done.

The linear prediction coefficient storage unit 241 is input for each frame when the missing determination information of the frame indicates that the sound signal code is not missing, that is, when the sound signal code of the frame is not missing. The decoded linear prediction coefficients ^ α ₁ , ^ α ₂ , ..., ^ α _p are stored as the preceding frame decoded linear prediction coefficients ^ β ₁ , ^ β ₂ , ..., ^ β _p .

Further, the linear prediction coefficient storage unit 241 stores the loss determination information of the frame when the loss signal information indicates that the sound signal code is missing, that is, when the sound signal code of the frame is missing. Output the preceding frame decoded linear prediction coefficients ^ β ₁ , ^ β ₂ , ..., ^ β _p to the linear prediction unit 243.

<Decoded sound signal storage unit 242>
The decoded sound signal storage unit 242 stores the loss determination information output by the non-packetizing unit 21 and the decoded sound signals ^ x (0), ^ x (1), ..., ^ x (N) output by the signal decoding unit 23. -1) is input.

  The decoded sound signal storage unit 242 stores decoded sound signals that have been successfully decoded for several frames, for example, two frames. For example, the decoded sound signal storage unit 242 stores the decoded sound signal when the missing signal determination information of the frame indicates that the sound signal code is not missing, that is, when the sound signal code of the frame is not missing. Each frame sample value of the decoded sound signal ^ y (N + 1), ^ y (N + 2),..., ^ Y (2N-1) immediately before the frame stored in the unit 242 is decoded two frames earlier. The sound signals ^ y (0), ^ y (1), ..., ^ y (N-1) are stored as respective sample values, and the input decoded sound signals ^ x (0), ^ x (1), ..., ^ x (N-1) are used as the sample values of the preceding frame decoded sound signal ^ y (N + 1), ^ y (N + 2), ..., ^ y (2N-1), respectively. Remember.

  Further, the decoded sound signal storage unit 242 is stored when the missing signal determination information of the frame indicates that the sound signal code is missing, that is, when the sound signal code of the frame is missing. The preceding frame decoded sound signal ^ y (0), ^ y (1), ..., ^ y (2N-1) is output to the linear prediction unit 243 and the signal extension selection unit 244.

<Linear prediction unit 243>
The linear predictor 243 includes, in the missing determination information, a frame indicating that the sound signal code is missing, that is, the missing portion output by the non-packeting unit 21 for the frame in which the sound signal code of the frame is missing. The determination information, the previous frame decoded linear prediction coefficients ^ β ₁ , ^ β ₂ , ..., ^ β _p output by the linear prediction coefficient storage unit 241 and the previous frame decoded sound signal ^ y output by the decoded sound signal storage unit 242 (0), ^ y (1), ..., ^ y (2N-1) are input.

ただし、n=0,1,…,N-1で、k>0においてpredict(-k)=^y(2N-k)である。 When the missing determination information of the frame indicates that the sound signal code is missing, that is, when the sound signal code of the frame is missing, the linear prediction unit 243 determines whether or not the previous frame decoded linear prediction coefficient ^ β _{Based on} the values of ₁ , ^ β ₂ , ..., ^ β _p , predictive signals predict (0), predict (1), ..., predict (N-1) are generated as in the following equation (1), for example, The generated prediction signals predict (0), predict (1),..., Predict (N−1) are output to the interpolation signal combining unit 245 (step D4).

Where n = 0, 1,..., N−1, and predict (−k) == y (2N−k) when k> 0.

  That is, the predictive signal predict (0), predict (1),..., Predict (N-1) defines the decoded linear prediction coefficient of the previous frame in which the sound signal code is not missing as the decoded linear prediction coefficient of the frame. This is a predicted signal of the frame when 0 is used as each sample value of the residual signal of the frame.

<Signal extension selection unit 244>
The signal extension selection unit 244 includes, in the loss determination information, a frame indicating that the sound signal code is missing, that is, the loss determination information output by the non-packetizing unit 21 for the frame in which the sound signal code is missing. , Yy (0), ^ y (1),..., ^ Y (2N-1) output from the decoded sound signal storage unit 242.

  As will be described later, the polarity of the preceding frame decoded sound signal ^ y (0), ^ y (1),..., ^ Y (2N-1) is reversed in time and the polarity is reversed in time. Each of them is, under certain assumptions, equal to a phase shifted by the frame length while maintaining the power spectrum of the decoded sound signal of the previous frame. Therefore, the signal extension selecting unit 244 determines, for each frame, when the loss determination information of the frame indicates that the sound signal code is missing, that is, when the sound signal code of the frame is missing, An appropriate one as a decoded sound signal of the frame (hereinafter, sometimes referred to as “current frame”) assumed from the decoded sound signal of the previous frame of the frame is selected as follows, for example, and the assumption is made. The extended decoded sound signal extend (0), extend (1),..., Extend (2N-1), which is the decoded sound signal of the current frame, is generated according to the above (step D5).

  The decoded sound signal of the previous frame ^ y (0), ^ y (1), ..., ^ y (2N-1) is, for example, the addition of a sine wave that is even symmetric about the boundary between the Nth and N + 1th samples. If it is expressed as a combination, the decoded sound signal of the current frame following that signal is the decoded sound signal of the previous frame ^ y (0), ^ y (1), ..., ^ y (2N-1). It becomes equal to what was arranged in reverse. Also, the decoded sound signal of the previous frame ^ y (0), ^ y (1), ..., ^ y (2N-1) is, for example, a sine wave that is oddly symmetric about the boundary between the Nth and N + 1th samples. , The decoded sound signal of the current frame following that signal is the decoded sound signal of the previous frame ^ y (0), ^ y (1), ..., ^ y (2N-1) Are reversed in time, and are equal to those obtained by inverting the polarity.

From this, first, the signal extension selecting unit 244 selects the Nth and N + 1th samples of the decoded sound signal of the previous frame ^ y (0), ^ y (1), ..., ^ y (2N-1). The sum and difference energy even and odd of the sample pair symmetrical about the boundary are calculated as in the following equations (2) and (2 '), and based on the magnitude of the value, the decoded sound signal of the previous frame ^ Determine whether y (0), ^ y (1), ..., ^ y (2N-1) is even or odd symmetric.

  Then, when even ≧ odd, the signal extension selecting unit 244 determines that the extended decoded sound signal is extend (n) = ^ y (2N-1-n) (n = 0,1) which is the reverse order of the decoded sound signal of the previous frame. , ..., 2N-1), and in the case of odd> even, extend (n) =-^ y (2N-1-n) (n = 0,1) which is the reverse order of the decoded sound signal of the previous frame whose polarity is inverted. , ..., 2N-1) and outputs the selected extended decoded sound signal extend (0), extend (1), ..., extend (2N-1) to the interpolation signal combining unit 245.

  That is, the interpolation signal generation unit 24, more specifically, the signal extension selection unit 244 of the interpolation signal generation unit 24 samples the decoded sound signal sample of the previous frame when the decoded sound signal of the previous frame is even symmetric. Select the signals arranged in reverse time, and if the decoded sound signal of the previous frame is odd-symmetric, select the signal whose polarity is inverted, and convert the selected signal to the extended decoded sound of the frame. Signal.

  In this way, the interpolation signal generation unit 24, more specifically, the signal extension selection unit 244 of the interpolation signal generation unit 24, for a frame in which the sound signal code is missing, decodes the decoded sound signal of the previous frame of the frame. Or a signal in which the polarity of the signal is inverted in time, or a signal in which the polarity of the signal is inverted, is selected as an extended decoded sound signal of the frame.

<Interpolation signal combining unit 245>
The interpolation signal combining unit 245 includes the loss determination information output from the depacketizing unit 21 and the decoded sound signals ^ x (0), ^ x (1), ..., ^ x (N- 1), the prediction signals predict (0), predict (1),..., Predict (N−1) output from the linear prediction unit 243, and the extended decoded sound signal extend (0), output from the signal extension selection unit 244. extend (1), ..., extend (2N-1) is input.

  The interpolation signal combining unit 245 determines whether the sound signal code of the frame is missing and the sound signal code of the previous frame according to the missing determination information of the frame and the previous frame. Then, for example, the following three processes are selectively performed.

  The interpolation signal combining unit 245 determines that both the missing signal determination information of the frame and the missing determination information of the previous frame indicate that the sound signal code is not missing, that is, the sound signal code of the frame is not missing. If the sound signal code of the previous frame of the frame is not missing, the decoded sound signal ^ x (0), ^ x (1), ..., ^ x (N-1) is converted to the decoded sound signal ^ X of the frame. Output as (0), ^ X (1), ..., ^ X (N-1) (step D6).

The interpolation signal combining unit 245, when the loss determination information of the frame indicates that the sound signal code is missing, that is, when the sound signal code of the frame is missing, for example, the following equation (3) Thus, the prediction signal predict (0), predict (1),..., Predict (N-1) and the extended decoded sound signal extend (0), extend (1),. The linearly combined signal using the window function is output as decoded sound signals ^ X (0), ^ X (1), ..., ^ X (N-1) of the frame (step D6).

  Here, n = 0, 1,..., N−1, and w (n) indicates a window function. As the window function w (n), a window function other than the above equation may be used.However, the window function by which extend (n) is multiplied with time, that is, the larger the value of n, the larger the value of n. It is desirable that the window function that multiplies predict (n) be a function that decreases with time, that is, a function that becomes smaller as the value of n increases. In this way, the linear combination of two types of signals using a window function is called a crossfade here, and a function for crossfading a window function multiplying each of extend (n) and predict (n) is called here. .

  In this way, the interpolation signal generation unit 24, more specifically, the interpolation signal combination unit 245 of the interpolation signal generation unit 24, for a frame in which the sound signal code is missing, decodes the decoded sound signal of the previous frame of the frame. A signal generated based on an extended decoded sound signal that is a signal obtained by arranging the samples in reverse in time or a signal obtained by inverting the polarity of the signal, and a signal obtained by performing linear prediction synthesis from the previous frame, decodes the frame. Sound signal.

The interpolation signal coupling unit 245 indicates that the missing signal determination information of the frame indicates that the sound signal code is not missing and the missing determination information of the previous frame of the frame indicates that the sound signal code is missing, that is, If the sound signal code of the frame is not missing and the sound signal code of the previous frame of the frame is missing, for example, the decoded sound signal ^ x (0), Cross-fading ^ x (1), ..., ^ x (N-1) and extended decoded sound signal extend (N), extend (N + 1), ..., extend (2N-1) Are output as decoded sound signals ^ X (0), ^ X (1), ..., ^ X (N-1) (step D6).

  Here, n = 0, 1,..., N−1, and w (n) indicates the above-described window function. By this operation, continuity between the frame in which information is missing and the current frame can be increased.

<Interpolation signal generator 24>
The interpolation signal generation unit 24 uses the processing of the linear prediction coefficient storage unit 241, the decoded sound signal storage unit 242, the linear prediction unit 243, the signal extension selection unit 244, and the equation (3) of the interpolation signal combination unit 245 described above. As for the frame in which the sound signal code is missing, the continuity with the previous frame is temporally selected from among a plurality of signal candidates having the same power spectrum as the decoded sound signal of the previous frame of the frame. It can be said that a candidate for a high signal is selected and the candidate for the selected signal is used as a decoded sound signal of the frame.

  When the sound signal code of the frame is missing, the interpolation signal combining unit 245 of the interpolation signal generation unit 24 uses the extended decoded sound signal output by the signal extension selecting unit 244 as the decoded sound signal of the frame. Is also good. That is, the interpolation signal generation unit 24, for a frame in which a sound signal code is missing, a signal obtained by arranging samples of the decoded sound signal of the previous frame of the frame in time reverse or the polarity of the signal is inverted. May be used as the decoded sound signal of the frame. In this case, the interpolation signal generation unit 24 may not operate the linear prediction coefficient storage unit 241 and the linear prediction unit 243, and may not include the linear prediction coefficient storage unit 241 and the linear prediction unit 243.

  In this way, the decoding apparatus performs interpolation using the same signal on the power spectrum of the previous frame based on only the information of the previous frame, so that there is no need to transmit additional information, and the decoding apparatus performs the same processing as normal decoding processing. With a delay within the range of the principle delay, it is possible to obtain a decoded sound signal that is more audibly better than the prior art.

[Decoding device and method according to second embodiment]
The decoding device and method according to the second embodiment realize a highly continuous crossfade by using a prediction signal in selecting a signal expansion method in the signal expansion selection unit 244. Hereinafter, details of the decoding device and method of the second embodiment will be described.

  FIG. 6 shows an example of the decoding device of the second embodiment. As in the first embodiment, the decoding device of the second embodiment includes, for example, a non-packetizing unit 21, a linear prediction coefficient decoding unit 22, a signal decoding unit 23, and an interpolation signal generation unit 24. The interpolation signal generation unit 24 includes, for example, a linear prediction coefficient storage unit 241, a decoded sound signal storage unit 242, a linear prediction unit 243, a signal extension selection unit 244, and an interpolation signal combination unit 245.

  Hereinafter, the linear prediction unit 243 and the signal extension selection unit 244 that are different from the first embodiment will be described. The same parts as in the first embodiment will not be described repeatedly.

<Linear prediction unit 243>
The linear prediction unit 243 includes the loss determination information output from the depacketizing unit 21 and the previous frame decoded linear prediction coefficients ^ β ₁ , ^ β ₂ , ..., ^ β _p output from the linear prediction coefficient storage unit 241. The preceding frame decoded sound signal ^ y (0), ^ y (1), ..., ^ y (2N-1) output from the decoded sound signal storage unit 242 is input.

  When the sound signal code is missing, the linear prediction unit 243 converts the prediction signals predict (0), predict (1),..., Predict (N-1) in the same manner as the linear prediction unit 243 of the first embodiment. Generate (Step D4).

  The linear prediction unit 243 outputs the generated prediction signals predict (0), predict (1),..., Predict (N−1) to the interpolation signal combining unit 245 and the signal extension selecting unit 244.

  The prediction signal predict (0), predict (1),..., Predict (N-1) is also different from that of the first embodiment in that it is also output to the signal extension selection unit 244.

<Signal extension selection unit 244>
The signal extension selection unit 244 includes, in the loss determination information, a frame indicating that the sound signal code is missing, that is, the loss determination information output by the non-packetizing unit 21 for the frame in which the sound signal code is missing. , Yy (0), ^ y (1), ..., ^ y (2N-1) output from the decoded sound signal storage unit 242, and the prediction signal predict ( 0), predict (1), ..., predict (N-1) are input.

  The signal extension selection unit 244 may determine whether the missing sound signal code of the frame is missing, that is, if the missing sound signal code of the frame is missing, for example, as follows. The signals extend (0), extend (1), ..., extend (2N-1) are selected (step D5).

ただし、n=0,1,…,2N-1である。 The signal extension selection unit 244 includes, for example, the following four candidates extend ₁ (n), extend ₂ (n) as candidates for the extended decoded sound signal extend (0), extend (1),..., Extend (2N-1). ), extend ₃ (n) and extend ₄ (n) are prepared. extend ₁ (n) is a signal obtained by arranging the preceding frame decoded sound signals ^ y (0), ^ y (1), ..., ^ y (2N-1) in reverse time. extend ₂ (n) is a signal obtained by inverting the polarity of the signal obtained by arranging the decoded sound signal of the previous frame ^ y (0), ^ y (1), ..., ^ y (2N-1) in reverse time. is there. extend ₃ (n) is the previous frame decoded sound signal ^ y (0), ^ y (1), ..., ^ y (2N-1) itself. extend ₄ (n) is a signal obtained by inverting the polarity of the preceding frame decoded sound signal ^ y (0), ^ y (1), ..., ^ y (2N-1).

Here, n = 0, 1,..., 2N−1.

  As described below, these candidates are equivalent to those obtained by shifting the phase by the frame length while maintaining the power spectrum of the decoded sound signal of the previous frame under a specific assumption. The signal extension selection unit 244 determines, for each frame, when the missing determination information of the frame indicates that the sound signal code is missing, that is, when the sound signal code is missing, In this case, the continuity between the decoded sound signal of the previous frame and the extended decoded sound signal is used as a reference. Then, the signal extension selecting unit 244 uses the prediction signal as a criterion of the continuity. Since the prediction signal is a signal continuous with the decoded sound signal of the previous frame on the principle of linear prediction, the signal extension selecting unit 244 determines the closeness of the value between the prediction signal and the candidate of the extended decoded sound signal as a criterion of continuity. The evaluation is performed using

ただし、i=1,2,3,4である。 That is, the signal extension selecting unit 244 selects, for example, the following equation (5) among the _four candidates extend ₁ (n), extend ₂ (n), extend ₃ (n), and extend ₄ (n). The extended decoded sound signal extend (0), extend (1),..., Extend (2N-1) is selected as a signal having the minimum square distance between the predicted signal and the extended decoded sound signal candidate. The extended decoded sound signals extend (0), extend (1),..., Extend (2N−1) are output to the interpolation signal combining unit 245.

Here, i = 1, 2, 3, and 4.

ただし、i=1,2,3,4である。 Alternatively, the signal extension selection unit 244 selects one of the above four candidates, extend ₁ (n), extend ₂ (n), extend ₃ (n), and extend ₄ (n), for example, as shown in the following equation (6). The signal with the largest inner product value may be selected as the extended decoded sound signal.

Here, i = 1, 2, 3, and 4.

In Equations (5) and (6), a signal extend _i (n) ′ obtained by multiplying extend _i (n) by the value of the window function for crossfading is expressed by Equations (5) and (6). ) May be used instead of extend _i (n). Similarly, in formulas (5) and (6), a signal predict (n) ′ obtained by multiplying predict (n) by the value of the window function for performing the crossfading is expressed by the above formulas (5) and (6). ) May be used instead of predict (n).

  In this manner, the interpolation signal generation unit 24, more specifically, the signal extension selection unit 244 of the interpolation signal generation unit 24 outputs a signal obtained by arranging the samples of the decoded sound signal of the previous frame of the current frame in reverse time, Among the signal in which the polarity of the signal is inverted, the decoded sound signal of the previous frame of the frame, and the signal of which the polarity of the decoded sound signal of the previous frame of the frame is inverted, similarity to the signal obtained by linear prediction synthesis from the previous frame. Is selected, and the selected signal is used as an extended decoded sound signal.

  Further, in this manner, the interpolation signal generation unit 24, more specifically, the signal extension selection unit 244 of the interpolation signal generation unit 24 decodes the previous frame of the frame in which the sound signal code is missing for the frame in which the sound signal code is missing. Any one of a signal obtained by arranging samples of the sound signal in reverse time, a signal in which the polarity of the signal is inverted, a decoded sound signal in the previous frame of the frame, and a signal in which the polarity of the decoded sound signal in the previous frame of the frame is inverted. Is selected, and the selected signal is used as an extended decoded sound signal.

  The signal extension selecting unit 244 outputs the selected extended decoded sound signal extend (0), extend (1),..., Extend (2N−1) to the interpolation signal combining unit 245.

  In the first embodiment, a signal in which the preceding frame decoded sound signal ^ y (0), ^ y (1),..., ^ Y (2N-1) is temporally reversed and a signal in which the polarity of the signal is inverted Are candidates for the extended decoded sound signal extend (0), extend (1),..., Extend (2N-1). On the other hand, in the second embodiment described above, the number of candidates for the extended decoded sound signal extend (0), extend (1),..., Extend (2N−1) is four. By increasing the number of candidates for the extended decoded sound signal extend (0), extend (1),..., Extend (2N−1), more accurate interpolation becomes possible.

  Note that also in the second embodiment, similarly to the first embodiment, the decoded sound signals of the previous frame ^ y (0), ^ y (1), ..., ^ y (2N-1) are temporally reversed. , Extend (0), extend (1),..., Extend (2N−1).

In this case, the signal extension selecting unit 244 calculates, for example, the sum and difference energies even and odd of the samples symmetric about the boundary when the decoded sound signal of the previous frame and the prediction signal are aligned, as defined by the following equation. Ask for each.

  Then, for the same reason as in the first embodiment, the signal extension selecting unit 244 sets extend (n) = ^ y (2N−1), which is the reverse order of the decoded sound signal of the previous frame, as the extended decoded sound signal when even ≧ odd. -n) (n = 0,1, ..., 2N-1), and in the case of odd> even, extend (n) =-^ y (2N-1) which is the reverse order of the decoded sound signal of the previous frame whose polarity has been inverted. -n) (n = 0, 1,..., 2N−1), and the selected extended decoded sound signals extend (0), extend (1),. Output to

[Decoding device and method according to third embodiment]
The decoding device and method of the third embodiment are for estimating a linear prediction coefficient based on a decoded sound signal of a frame before a frame in which information is lost. Thereby, even when the encoding device assumed by the decoding device does not include the linear prediction analysis unit 11 and the packet input to the decoding device does not include the linear prediction coefficient code, Signal interpolation using linear prediction can be performed. Alternatively, the decoding device can perform higher-precision prediction by obtaining the order of the linear prediction analysis higher than the order used in the encoding.

  FIG. 7 shows an example of a decoding device according to the third embodiment. The decoding device according to the third embodiment includes, for example, a non-packetizing unit 21, a linear prediction coefficient decoding unit 22, a signal decoding unit 23, and an interpolation signal generation unit 24. The interpolation signal generation unit 24 includes, for example, a decoded sound signal storage unit 242, a linear prediction unit 243, a signal extension selection unit 244, an interpolation signal combination unit 245, and a linear prediction coefficient estimation unit 246.

  Hereinafter, a description will be given of the non-packetizing unit 21, the linear prediction coefficient decoding unit 22, the decoded sound signal storage unit 242, the linear prediction unit 243, and the linear prediction coefficient estimation unit 246, which are different from the first embodiment or the second embodiment. I do. The same parts as those in the first embodiment or the second embodiment will not be described repeatedly.

<Non-packetizing unit 21>
The non-packetizing unit 21 generates missing judgment information by the same processing as in the first embodiment or the second embodiment, and stores the generated missing judgment information in the decoded sound signal storage unit 242, the signal extension selecting unit 244, and The data is output not only to the interpolation signal combining unit 245 but also to the linear prediction coefficient estimating unit 246 (step D1).

  Other processes of the depacketizing unit 21 are the same as those of the first embodiment or the second embodiment.

<Linear prediction coefficient decoding unit 22>
The linear prediction coefficient decoding unit 22 performs the decoding linear prediction coefficients ^ α ₁ , ^ α ₂ , ... to give the ^ alpha _p, resulting decoded linear prediction coefficient _{^ α 1, ^ α 2,} ..., and outputs a ^ alpha _p to the signal decoding unit 23 (step D2).

  Other processes of the linear prediction coefficient decoding unit 22 are the same as those of the first embodiment or the second embodiment.

<Decoded sound signal storage unit 242>
If the missing signal determination information of the frame indicates that the sound signal code is missing, that is, if the sound signal code of the frame is missing, the decoded sound signal storage unit 242 stores the stored previous frame. The decoded sound signal ^ y (0), ^ y (1), ..., ^ y (2N-1) is output not only to the linear prediction unit 243 and the signal extension selection unit 244, but also to the linear prediction coefficient estimation unit 246.

  Other processes of the decoded sound signal storage unit 242 are the same as those of the first embodiment or the second embodiment.

<Linear prediction coefficient estimation unit 246>
The linear prediction coefficient estimating unit 246 includes the loss determination information output by the depacketizing unit 21 and the decoded sound signal of the previous frame ^ y (0), ^ y (1), ..., output by the decoded sound signal storage unit 242. ^ y (2N-1) is input.

The linear prediction coefficient estimating unit 246 outputs the preceding frame decoded sound signal ^ when the loss determination information of the frame indicates that the sound signal code is missing, that is, when the sound signal code of the frame is missing. Based on y (0), ^ y (1),..., ^ y (2N-1), the estimated linear prediction coefficient ^ γ ₁ , ^ γ ₂ , ..., ^ γ _p are generated (step D7).

Linear predictive coefficient estimator 246, generated estimated linear prediction coefficient _{^ γ 1, ^ γ 2,} ..., and outputs a ^ gamma _p the linear prediction unit 243.

<Linear prediction unit 243>
The linear predictor 243 includes the missing information output from the depacketizer 21, the estimated linear prediction coefficients ^ γ ₁ , ^ γ ₂ , ..., ^ γ _p output from the linear prediction coefficient estimator 246, and the decoded sound. The previous frame decoded sound signal ^ y (0), ^ y (1), ..., ^ y (2N-1) output from the signal storage unit 242 is input.

The linear prediction unit 243 determines, for each frame, the estimated linear prediction coefficient when the missing signal determination information of the frame indicates that the sound signal code is missing, that is, when the sound signal code of the frame is missing. Based on the values of ^ γ ₁ , ^ γ ₂ ,…, ^ γ _p , predictive signals predict (0), predict (1),…, predict (N-1) are calculated as in the following equation (1 '), for example. The generated prediction signals predict (0), predict (1),..., Predict (N−1) are output to the interpolation signal combining unit 245.

Here, n = 1, 2,..., N. The prediction signal predict (0), predict (1 ), ..., predict (N-1) is estimated linear prediction coefficient _{^ γ 1, ^ γ 2,} ..., a ^ gamma _p is the decoded linear prediction coefficients of the frame, This is a predicted signal of the frame when 0 is used as each sample value of the residual signal of the frame.

[Technical background]
In the above embodiment, interpolation of a frame with missing information is performed using at least one of the two types of signals.

The first of the two types of signals is a prediction signal predict (0), predict (1),..., Predict (N-1). Prediction signal predict (0), predict (1 ), ..., predict (N-1) is decoded linear prediction coefficients of the previous frame _{^ β 1, ^ β 2,} ..., the formula using the ^ β _p (1) For example, it is required. When the prediction signals predict (0), predict (1),..., Predict (N−1) are used, continuity with the previous frame is ensured due to its property. If it is short, the energy of the prediction signal gradually decreases, and the value of the prediction signal becomes 0 in the latter half of the frame. Therefore, if interpolation of a frame with missing information is performed using only the prediction signal, discontinuity with the next frame may occur.

The decoded linear prediction coefficients ^ β ₁ , ^ β ₂ , ..., ^ β _p are obtained by decoding on the decoding side what has been encoded on the encoding side, as in the first and second embodiments. be able to.

Note that, instead of the decoded linear prediction coefficients ^ β ₁ , ^ β ₂ ,..., ^ Β _p , the estimated linear prediction coefficients ^ γ obtained by linear prediction analysis from previously decoded waveforms as in the third embodiment. _1, ^ γ _2, ..., with ^ gamma _p, the prediction signal predict (0), predict (1 ), ..., can also be determined predict (N-1). In this case, p can be made as large as, for example, N, so that the energy of the prediction signal predict (0), predict (1),..., Predict (N-1) can be maintained up to a large number of samples. However, in order to obtain a linear prediction coefficient of a large order, it is necessary to hold a long past sample (a plurality of frames), and the amount of calculation for analysis becomes extremely large.

  The extended decoded sound signal, which is the second of the above two types of signals, is used to prevent loss of signals in the frame while preventing an increase in the amount of calculation. The extended decoded sound signal extend (0), extend (1),..., Extend (2N-1) includes the decoded sound signal ^ y (0), ^ y (1),. For example, a signal having the power spectrum of -1) is used. The extended decoded sound signal extend (0), extend (1),..., Extend (2N-1) and the prediction signals predict (0), predict (1),. By making the cross-fade signal as the decoded sound signal of the missing frame as in 3), the energy of the interpolated signal is maintained while maintaining the temporal and frequency continuity with the frames before and after the missing frame. be able to.

  Hereinafter, an example of selection of the extended decoded sound signal extend (0), extend (1),..., Extend (2N-1) will be described.

When considering the frequency of a signal, it is widely used to decompose the signal into sine waves, but there are various types of sine waves used in the decomposition. For example, as shown in FIG. 8, consider a case where a signal is decomposed by a cos wave at n = 0, 1,..., N−1 according to the following equation.

Extending the signal to n + N = N, ..., 2N-1 while maintaining the frequency spectrum a (0), a (1), ..., a (N-1) can be expressed as the following equation .

  That is, a signal obtained by reversing the signal in time is equivalent to a signal obtained by shifting the phase by the frame length while maintaining the power spectrum of the signal.

Similarly, when a certain signal is decomposed with a sine wave at n = 0,1, ..., N-1 according to the following equation, it is expanded to n + N = N, ..., 2N-1 as the following equation it can.

  In other words, a signal obtained by inverting the signal in time and inverting the polarity is equivalent to a signal obtained by shifting the phase by the frame length while maintaining the power spectrum of the signal.

  Due to the above signal extension, if the signal is assumed to be composed of a cosine wave, the signal is inverted in time, whereas if the signal is assumed to be composed of a sine wave, the signal is temporally reversed. It can be seen that a polarity-reversed one is obtained. Therefore, in the first embodiment, it is compared whether the even symmetry or the odd symmetry of the signal is stronger. If the even symmetry is stronger, it is considered that the signal is composed of an even-symmetry cos wave, and the odd symmetry is stronger. In this case, the extension signal is selected on the assumption that the signal is composed of sine waves having odd symmetry.

When a signal is decomposed by a complex sine wave, a signal expanded to n + N = N,..., 2N−1 can be obtained as in the following equation.

These are a signal as it is and a signal whose polarity is inverted, respectively. The above four extension methods correspond to extended decoded sound signal candidates extend ₁ (n), extend ₂ (n), extend ₃ (n), and extend ₄ (n) in the signal extension selecting unit 244 of the second embodiment. I do.

[Program and recording medium]
When the processing in each unit of the decoding device is realized by a computer, the processing content of the function that each unit of the decoding device should have is described by a program. By executing this program on a computer, the processing of each unit is realized on the computer.

  A program describing this processing content can be recorded on a computer-readable recording medium. As a computer-readable recording medium, for example, any recording medium such as a magnetic recording device, an optical disk, a magneto-optical recording medium, and a semiconductor memory may be used.

  Further, the processing of each unit may be configured by executing a predetermined program on a computer, or at least a part of these processing may be realized by hardware.

[Modification]
In addition, it goes without saying that changes can be made as appropriate without departing from the spirit of the present invention.

Claims (14)

A decoding device for obtaining a decoded sound signal for each frame,
For a frame in which the sound signal code is missing, the signal obtained by arranging samples of the decoded sound signal of the previous frame of the frame in reverse time and the signal in which the polarity of the signal is inverted are selected from the time of the previous frame. A signal generated based on an extended decoded sound signal that is a candidate for the selected signal and a signal that is linearly predicted and synthesized from the previous frame, and decodes the frame. An interpolated signal generation unit to be a sound signal,
A decoding device including: A decoding device for obtaining a decoded sound signal for each frame,
For a frame in which the sound signal code is missing, an extended decoded sound signal that is a signal obtained by arranging samples of the decoded sound signal of the previous frame of the above frame in time or a signal obtained by inverting the polarity of the signal, A signal generated based on a signal subjected to linear prediction synthesis from the previous frame, and an interpolation signal generation unit that serves as a decoded sound signal of the frame;
A decoding device including: A decoding device for obtaining a decoded sound signal for each frame,
For a frame in which the sound signal code is missing, a signal in which samples of the decoded sound signal of the previous frame of the frame are arranged in time reverse, a signal in which the polarity of the signal is inverted, the decoded sound signal of the previous frame, and From among the signals obtained by inverting the polarity of the decoded sound signal of the previous frame, a signal candidate having a high similarity to the signal obtained by linearly predicting and synthesizing the previous frame is selected, and the extended decoded sound signal which is the selected signal candidate is selected. Or, a signal generated based on the extended decoded sound signal and a signal obtained by performing linear prediction synthesis from the previous frame, and an interpolation signal generation unit that sets the decoded sound signal of the frame as a decoded sound signal.
A decoding device including: A decoding device for obtaining a decoded sound signal for each frame,
For a frame in which the sound signal code is missing, linear prediction synthesis was performed from the previous frame from a signal in which samples of the decoded sound signal of the previous frame of the frame were arranged in time reverse or a signal in which the polarity of the signal was inverted. A signal having a high similarity to the signal is selected, and an extended decoded sound signal that is the selected signal, or a signal generated based on the extended decoded sound signal and a signal that is linearly predicted and synthesized from the previous frame, An interpolated signal generation unit for the decoded sound signal of
A decoding device including: A decoding device for obtaining a decoded sound signal for each frame,
For a frame in which the sound signal code is missing, from among a plurality of signal candidates having the same power spectrum as the decoded sound signal of the previous frame of the frame, a signal having a high temporal continuity with the previous frame is selected. An interpolation signal generation unit that selects a candidate, and sets a signal generated based on an extended decoded sound signal that is a candidate for the selected signal and a signal that is linearly predicted and synthesized from the previous frame as a decoded sound signal of the frame; ,
A decoding device including: A decoding device for obtaining a decoded sound signal for each frame,
For a frame in which the sound signal code is missing, among a plurality of signal candidates having the same power spectrum as the decoded sound signal of the previous frame of the frame, the similarity to the signal obtained by linear prediction synthesis from the previous frame is similar. A candidate for a high signal is selected, and an extended decoded sound signal that is a candidate for the selected signal, or a signal generated based on the extended decoded sound signal and a signal that is linearly predicted and synthesized from the previous frame, An interpolated signal generator for decoding sound signals,
A decoding device including: The decoding device according to any one of claims 1 , 2 , and 4 ,
When the decoded sound signal of the previous frame is even symmetric, the interpolation signal generation unit selects a signal in which samples of the decoded sound signal of the previous frame are arranged in reverse time, and decodes the previous frame. If the sound signal is oddly symmetric, select the signal whose polarity is inverted, and use the selected signal as the decoded sound signal of the frame or the extended decoded sound signal,
Decoding device. A decoding method for obtaining a decoded sound signal for each frame,
For a frame in which the sound signal code is missing, the signal obtained by arranging samples of the decoded sound signal of the previous frame of the frame in reverse time and the signal in which the polarity of the signal is inverted are selected from the time of the previous frame. A signal generated based on an extended decoded sound signal that is a candidate for the selected signal and a signal that is linearly predicted and synthesized from the previous frame, and decodes the frame. Generating an interpolation signal as a sound signal,
A decoding method including: A decoding method for obtaining a decoded sound signal for each frame,
For a frame in which the sound signal code is missing, an extended decoded sound signal that is a signal obtained by arranging samples of the decoded sound signal of the previous frame of the above frame in time or a signal obtained by inverting the polarity of the signal, A signal generated based on a signal obtained by linear prediction synthesis from the previous frame, and an interpolation signal generation step of setting a signal generated as a decoded sound signal of the frame,
A decoding method including: A decoding method for obtaining a decoded sound signal for each frame,
For a frame in which the sound signal code is missing, a signal in which samples of the decoded sound signal of the previous frame of the frame are arranged in time reverse, a signal in which the polarity of the signal is inverted, the decoded sound signal of the previous frame, and From among the signals obtained by inverting the polarity of the decoded sound signal of the previous frame, a signal candidate having a high similarity to the signal obtained by linearly predicting and synthesizing the previous frame is selected, and the extended decoded sound signal which is a candidate of the selected signal is selected. Or, a signal generated based on the extended decoded sound signal and a signal obtained by linear prediction synthesis from the previous frame, and an interpolation signal generation step of setting the decoded sound signal of the frame as a decoded sound signal;
A decoding method including: A decoding method for obtaining a decoded sound signal for each frame,
For a frame in which the sound signal code is missing, linear prediction synthesis was performed from the previous frame from a signal in which samples of the decoded sound signal of the previous frame of the frame were arranged in time reverse or a signal in which the polarity of the signal was inverted. A signal having a high similarity to the signal is selected, and an extended decoded sound signal that is the selected signal, or a signal generated based on the extended decoded sound signal and a signal that is linearly predicted and synthesized from the previous frame, An interpolation signal generation step as a decoded sound signal of
A decoding method including: A decoding method for obtaining a decoded sound signal for each frame,
For a frame in which the sound signal code is missing, from among a plurality of signal candidates having the same power spectrum as the decoded sound signal of the previous frame of the frame, a signal having a high temporal continuity with the previous frame is selected. An interpolation signal generating step of selecting a candidate, and setting a signal generated based on the extended decoded sound signal that is a candidate for the selected signal and a signal obtained by linear prediction synthesis from the previous frame as a decoded sound signal of the frame ,
A decoding method including: A decoding method for obtaining a decoded sound signal for each frame,
For a frame in which the sound signal code is missing, among a plurality of signal candidates having the same power spectrum as the decoded sound signal of the previous frame of the frame, the similarity to the signal obtained by linear prediction synthesis from the previous frame is similar. A candidate for a high signal is selected, and an extended decoded sound signal that is a candidate for the selected signal, or a signal generated based on the extended decoded sound signal and a signal that is linearly predicted and synthesized from the previous frame, An interpolation signal generating step to be a decoded sound signal,
A decoding method including:   A program for causing a computer to function as each unit of the decoding device according to any one of claims 1 to 7.

Images