JP3713288B2 - Speech decoder - Google Patents

Description

[0001]
[Industrial application fields]
The present invention relates to a speech decoding apparatus that receives speech encoding parameters and decodes and reproduces original speech.
[0002]
[Prior art]
Speech coding / decoding devices used in mobile communication systems such as automobile telephones, in addition to lowering bit rates, cause code errors when mobile terminals pass through places with low electric field strength, such as valleys in buildings. It is required to be able to reproduce good audio even when The speech synthesizer described in Japanese Patent Publication No. 58-49874 is an example of a conventional speech decoding apparatus proposed as a response to such a demand.
[0003]
FIG. 4 is a block diagram of this conventional speech decoding apparatus, where 1 is an input terminal, 2 is an output terminal, 3 is a regenerative repeater demodulation circuit, 4 is a synthesis circuit, 5 is detected at the time of instantaneous interruption and recovery from instantaneous interruption. An instantaneous interruption detecting circuit 6 for delaying, a delay memory circuit 6 for delay storing one set of analysis parameters, and a switch 7.
[0004]
In this conventional apparatus, the received signal input from the input terminal 1 is regeneratively relay demodulated by the regenerative relay demodulation circuit 3, converted into a digital signal analysis parameter, and output to the delay storage circuit 6 and the switch 7. When it is not at the time of instantaneous interruption, the instantaneous interruption detection circuit 5 detects that it is not an instantaneous interruption, and outputs this to the delay memory circuit 6 and the switch 7. Based on the signal from the instantaneous interruption detection circuit 5, the delay storage circuit 6 inputs the analysis parameter from the regenerative repeater demodulation circuit 3 and sequentially stores the delay. The switch 7 connects the regenerative repeater demodulation circuit 3 and the synthesis circuit 4 based on the signal from the instantaneous interruption detection circuit 5. The synthesis circuit 4 restores the audio signal based on the output signal of the regenerative repeater demodulation circuit 3 via the switch 7 and outputs it to the output terminal 2.
[0005]
When a momentary interruption occurs in the input signal from the input terminal 1, the momentary interruption detection circuit 5 detects it and outputs this information to the delay storage circuit 6 and the switch 7. The delay storage circuit 6 stops delaying and storing the input signal from the regenerative repeater demodulation circuit 3 based on the instantaneous interruption information from the instantaneous interruption detection circuit 5, and repeatedly outputs the stored contents to the switch 7 in order from the oldest. Since the delay storage circuit 6 has a storage capacity of one set of analysis parameters, the analysis parameter of the previous set is output to the switch 7. The switch 7 connects the delay storage circuit 6 and the synthesis circuit 4 based on the instantaneous interruption information from the instantaneous interruption detection circuit 5. The synthesis circuit 4 restores an audio signal based on the input signal from the delay storage circuit 6 and outputs it to the output terminal 2. When the instantaneous interruption is recovered, the instantaneous interruption detection circuit 5 detects the instantaneous interruption recovery time and outputs instantaneous interruption recovery information to the delay storage circuit 6 and the switch 7. At this time, the operation returns to the case where the above-mentioned momentary interruption has not occurred.
[0006]
As described above, in the conventional speech decoding apparatus, when an error occurs in the speech coding parameter due to the instantaneous interruption of the radio wave, the speech is reproduced based on a set of speech coding parameters input at normal time. Even when this occurs, it is possible to prevent major deterioration in sound quality. However, the detection capability of the instantaneous interruption detection circuit 5 has a limit, and any detection method may be used to cause erroneous detection. Therefore, when an error exceeding the detection capability of the instantaneous interruption detection circuit 5 occurs, the sound quality of the reproduced sound may be greatly deteriorated.
[0007]
[Problems to be solved by the invention]
As described above, since the detection capability of the instantaneous interruption detection circuit is limited in the conventional speech decoding apparatus, the instantaneous interruption may be erroneously detected, and it is possible to detect that there is no error even in a frame in which there is a code error due to the instantaneous interruption. There is sex. In such a case, the speech coding parameters of the frames that have not been detected despite the presence of code errors are stored, and the speech is reproduced based on the speech coding parameters, so that the quality of the decoded speech deteriorates. There was a problem.
An object of the present invention is to provide a speech decoding apparatus capable of preventing deterioration in sound quality of reproduced speech due to erroneous detection of speech coding parameter errors.
[0008]
[Means for Solving the Problems]
In order to solve the above-described problem, the present invention provides an audio decoding apparatus that reproduces an audio signal by performing decoding based on an audio encoding parameter input for each unit interval , and an error in the input audio encoding parameter. an error detection means for detecting, storing means for storing across multiple unit sections in which no error is detected by the error detection means at least one set of speech coding parameters to be the input, past read from the storage means Parameter reproduction means for reproducing a speech coding parameter based on speech coding parameters over a plurality of unit intervals of
For a unit section in which no error is detected by the error detection means, a speech signal is reproduced based on the input speech coding parameter, and for a unit section in which an error is detected by the error detection means, reproduction is performed by the parameter reproduction means. And a reproducing means for reproducing the audio signal based on the audio encoding parameter.
[0009]
More specifically, an interpolating unit for interpolating speech coding parameters over a plurality of past unit intervals read from the storage unit as the parameter reproducing unit is provided, and the unit interval in which the error is detected by the error detecting unit For the above, the audio signal may be reproduced based on the audio encoding parameter obtained by the interpolation means.
[0010]
Furthermore, in the present invention, there is provided selection means for selecting the most reliable speech coding parameter among speech coding parameters over a plurality of past unit intervals read from the storage means, and an error is detected by the error detection means. With respect to the unit section, the reproduction unit may reproduce the audio signal based on the audio encoding parameter selected by the selection unit.
[0011]
[Action]
In the speech decoding apparatus according to the present invention, the presence / absence of a code error in the speech coding parameter input by the error detection unit is detected, and in the unit interval in which no error is detected, the playback unit performs the playback based on the input speech coding parameter. together with the audio are reproduced, at least one set of the storage unit thus speech coding parameters are stored. In this way, the speech coding parameters of a plurality of unit sections in which no error is detected are stored in the storage means, and further, for example, interpolation calculation is performed on the past plurality of unit section encoding parameters read from the storage means. Done by means. Then, in the unit section where the error is detected, the voice is reproduced based on the voice coding parameter obtained by this interpolation means.
[0012]
Here, even if an error has occurred in the input speech coding parameter, even if it is detected that there is no error due to an error detection by the error detection means, such error detection continues over two or more unit intervals. The probability of occurrence is very low. Therefore, when performing the interpolation calculation of the speech coding parameters of a plurality of unit sections detected as error-free, even if the speech coding parameter of any unit section is incorrect, other unit sections in which no error has occurred As a result of alleviating the influence of the error by the speech coding parameter, the reliability of the speech coding parameter obtained by the interpolation calculation is improved, so that the speech is reproduced based on the speech coding parameter of only one unit section. Compared with the conventional method, the deterioration of sound quality is remarkably reduced.
[0013]
Further, the reliability of the speech coding parameters of a plurality of unit sections detected as having no error is determined by some means, and the speech is reproduced using the speech coding parameter with the highest reliability based on the determination result. Even if it carries out, the sound quality degradation of reproduction | regeneration audio | voice is reduced similarly.
[0014]
【Example】
Embodiments of the present invention will be described below with reference to the drawings.
(Example 1)
FIG. 1 is a block diagram of a speech decoding apparatus according to an embodiment of the present invention. The speech decoding apparatus according to this embodiment includes a coding parameter decoder 101 that decodes a speech coding parameter input to an input terminal 100, an error detection circuit 102 that detects an error of a speech coding parameter that is input, and a switch 103, first and second storage circuits 104 and 105 for storing speech coding parameters of a frame (unit section) detected as “no error” by the error detection circuit 102, and the storage circuits 104 and 105 An interpolation circuit 106 that performs an interpolation operation on the read speech coding parameters, a switch 107 that selects either the output of the coding parameter decoder 101 or the output of the interpolation circuit 106, and the speech selected by the switch 107 A speech synthesizer 108 that synthesizes and reproduces the original speech from the encoding parameters and outputs it to the output terminal 109. Constructed. The switch 103, the interpolation circuit 106, and the switch 107 are controlled by the output of the error detection circuit 102.
[0015]
Next, the operation of the speech decoding apparatus according to this embodiment will be described.
A plurality of speech encoding parameters transmitted from a transmitting side speech encoding apparatus (not shown) are input to the input terminal 100 in units of frames. For example, when the encoding method of the speech encoding apparatus is a CELP (Code Excited Linear Prediction) method, these speech encoding parameters are LPC analysis parameters, synthesis filter drive signals, speech signal pitches and gains. It is a parameter. Also, these speech coding parameters are transmitted after being error correction coded.
[0016]
These speech encoding parameters input to the input terminal 100 are decoded by the encoding parameter decoder 101 and further output to the switches 103 and 107.
[0017]
The speech coding parameters input to the input terminal 100 are also input to the error detection circuit 102, where it is detected whether there is a code error. The switches 103 and 107 are controlled by the output of the error detection circuit 102. That is, when the output of the error detection circuit 102 indicates “no error”, the switch 107 selects the speech coding parameter decoded by the coding parameter decoder 101 and sends it to the speech synthesizer 108. At this time, the switch 103 is turned on, and the speech coding parameter decoded by the coding parameter decoder 101 is sent to the first storage circuit 104. The first storage circuit 104 outputs the content stored and held so far to the second storage circuit 105, and then stores the new speech coding parameter input through the switch 103, and at the same time stores and holds it. The contents thus output are output to the interpolation circuit 106. On the other hand, the second storage circuit 105 clears the contents stored and held so far, and then stores the contents input from the first storage circuit 104.
[0018]
In this way, the first and second storage circuits 104 and 105 store and hold the speech coding parameters of the frames in which the code error is not detected by the error detection circuit 102 for two frames in the newest order.
[0019]
Next, when the output of the error detection circuit 102 indicates “with error”, the switch 103 is turned off, and the input of the speech coding parameter in which the error is detected to the storage circuit 104 is blocked. At the same time, the interpolation circuit 106 performs an interpolation calculation of speech coding parameters based on the contents of the first and second storage circuits 104 and 105. Specifically, for example, when an LSP parameter is used as an LPC analysis parameter that is one of speech coding parameters, the interpolation circuit 106 performs interpolation of the LSP parameter according to the following equation.
[0020]
ωi (n) = αωi (n−1) + (1−α) ωi (n−2) (1)
Here, 1 ≦ i ≦ P, P is the analysis order, ωi (n) is the LSP parameter of the current frame, and ωi (n−1) and ωi (n−2) are stored and held in the storage circuits 104 and 105, respectively. This is the first and second newest parameter among the LSP parameters of the frame detected as “no error”.
[0021]
Α is a coefficient of 0 <α <1 for weighted averaging of ωi (n−1) and ωi (n−2), and more realistically, for example, set in a range of 0.8 <α <1. Is done. Thus, the reason why the weighting coefficient α for ωi (n−1) is made larger than the weighting coefficient (α−1) for ωi (n−2) is that ωi (n−1) is actually the interpolation calculation result. This is because it is close in time to the required speech coding parameters.
[0022]
When the output of the error detection circuit 102 indicates “There is an error”, the speech coding parameter interpolated by the interpolation circuit 106 as described above is selected by the switch 107 and input to the speech synthesizer 108. The voice synthesizer 108 reproduces a voice signal based on the inputted voice coding parameter and outputs it to the output terminal 102.
[0023]
As described above, in this embodiment, for at least one set of speech coding parameters of frames in which errors are detected (LSP parameters in the above example), the parameters of two frames in which no errors have been detected in the past are weighted averaged. The interpolated parameters are substituted. Since there is a limit to the detection capability of the error detection circuit 102, there is a possibility of erroneous detection as described above. However, since the probability of repeated erroneous detection is very small, as in this embodiment. The reliability of the interpolated parameters is greatly improved by performing weighted averaging of the speech coding parameters of two frames, and the sound quality of the reproduced speech in a frame in which a code error has occurred by using this for decoding and reproducing speech signals. Deterioration can be prevented. Also, in the conventional method, speech encoding parameters are interpolated based on parameters of one frame, so in this embodiment, interpolation accuracy is improved because it is performed based on parameters of two frames. In addition, the effect of further improving the quality of the reproduced sound can be expected.
[0024]
In the above embodiment, the speech coding parameters of two frames are stored and held, and based on the contents, the erroneous frame parameters are obtained by secondary interpolation, but the number of storage circuits is increased, It may be configured to perform higher-order interpolation. In this case, it is possible to reduce the influence of erroneous detection of errors and to increase the interpolation accuracy.
[0025]
(Example 2)
FIG. 2 is a block diagram showing a configuration of a decoding apparatus according to another embodiment of the present invention. The same parts as those in the embodiment shown in FIG. 1 are denoted by the same reference numerals, and the differences will be mainly described. In this embodiment, the speech coding parameters are interpolated by moving average, which is different from the embodiment of FIG. different.
[0026]
That is, in FIG. 2, the speech encoding parameter decoded by the encoding parameter decoder 101 when the output of the error detection circuit 102 indicates “no error” is stored and held in the first storage circuit 104 . The interpolation circuit 202 includes coefficient multipliers 203 and 204, an adder 205, and a second storage circuit 206. The storage circuit 206 stores and holds the contents in units of frames in which the output of the error detection circuit 102 indicates “no error”. To be controlled. In this embodiment, the interpolated value of the speech coding parameter is calculated by the following equation.
[0027]
Ωi = βΩi + (1-β) ωi (n−1) (2)
Here, Ωi is the interpolation value of the LSP parameter, the output of the adder 205, ωi (n−1) is the LSP parameter of the frame in which the output of the error detection circuit 102 indicates “no error”, and (1-β) is The coefficient of the multiplier 203, β is the coefficient of the multiplier 204 (moving average coefficient).
[0028]
According to this embodiment, since the more moving average parameters of all frames without previous error interpolation speech coding parameters, reliability is improved and the interpolation accuracy of the interpolation parameter. In addition, since high-order interpolation is performed by the interpolation circuit 202 including two coefficient multipliers 203 and 204, an adder 205, and one storage circuit 206, a simple addition average as in the embodiment of FIG. Therefore, there is an advantage that the circuit scale can be reduced as compared with the case where higher-order interpolation is realized.
[0029]
In each of the above embodiments, the speech coding parameter interpolation calculation is performed in units of frames, but may be performed in units of subframes. In that case, since the correlation of parameters used for interpolation becomes strong, there is an effect of improving the interpolation accuracy.
[0030]
(Example 3)
FIG. 3 is a block diagram showing a configuration of a speech decoding apparatus according to still another embodiment of the present invention. The same parts as those in the embodiment shown in FIG. 1 are denoted by the same reference numerals, and the differences will be mainly described. In this embodiment, a reliability determination unit 301 and a switch 302 are provided, instead of performing speech coding parameter interpolation. 1 is different from the embodiment of FIG. 1 in that the audio encoding parameters stored and held in the first and second storage circuits 104 and 105 are selectively switched and used for decoding and reproducing audio signals.
[0031]
The first and second storage circuits 104 and 105 respectively store the speech coding parameters decoded by the coding parameter decoder 101 when the output of the error detection circuit 102 indicates “no error”. The reliability determination unit 301 determines that the continuity with the past speech encoding parameter is higher among the speech encoding parameters stored and held in the storage circuits 104 and 105 as being highly reliable, The switch 302 is controlled based on the determination result. When the error detection circuit 102 detects that “there is an error”, the speech encoding parameter selected by the switch 302 is selected by the switch 107, and based on this, the speech synthesizer 108 decodes and reproduces the speech.
[0032]
Note that when selecting speech coding parameters stored and held in the storage circuits 104 and 105, a filter included in the speech coding parameters is used in addition to a method that considers continuity with the past speech coding parameters described above. As for the coefficient, a speech coding parameter including a filter coefficient that makes the filter more stable may be determined as having higher reliability. Also, regarding the power coefficient included in the speech coding parameter, the speech coding parameter including the power coefficient that changes more smoothly from the power coefficient included in the past speech coding parameter is determined as having higher reliability. May be.
[0033]
As described above, according to the present embodiment, the speech coding parameter of the frame in which the error is detected is the speech coding parameter when the error stored in the plurality of storage circuits 104 and 105 is not detected. By selecting the parameter with the highest reliability and using it for audio reproduction, it is possible to prevent deterioration of the sound quality of the reproduced audio of a frame in which a code error has occurred.
[0034]
【The invention's effect】
As described above, according to the present invention, when there is an error in the speech coding parameter, the speech is reproduced and output based on the speech coding parameters of a plurality of unit sections determined as having no past errors. Therefore, it is possible to prevent deterioration in sound quality due to erroneous detection of error detection and to accurately perform speech coding parameter interpolation at the time of a code error, so that the effect of improving the reproduction sound quality can be obtained.
[Brief description of the drawings]
FIG. 1 is a block diagram showing a configuration of a speech decoding apparatus according to an embodiment of the present invention. FIG. 2 is a block diagram showing a configuration of a speech decoding apparatus according to another embodiment of the present invention. FIG. 4 is a block diagram showing the configuration of a speech decoding apparatus according to another embodiment. FIG. 4 is a block diagram showing the configuration of a conventional speech decoding apparatus.
DESCRIPTION OF SYMBOLS 100 ... Input terminal 101 ... Coding parameter decoder 102 ... Error detection circuit 103 ... Switch 104 ... 1st memory circuit 105 ... 2nd memory circuit 106 ... Interpolation circuit 107 ... Switch 108 ... Speech synthesizer 109 ... Output terminal 202 ... Interpolation circuit 203 ... Coefficient multiplier 204 ... Coefficient multiplier 205 ... Adder 206 ... Storage circuit 301 ... Reliability determination unit 302 ... Switch

Claims (1)

In a speech decoding device that reproduces a speech signal by performing decoding based on a speech coding parameter input for each unit section,
Error detection means for detecting an error in the input speech coding parameter;
Storage means for storing at least one set of the input speech coding parameters over a plurality of unit intervals in which no error is detected by the error detection means;
A speech coding parameter including a power coefficient in which a change from a power coefficient included in a past speech coding parameter changes more smoothly among speech coding parameters over a plurality of past unit intervals read from the storage unit. A selection means for selecting the most reliable speech coding parameter by determining that the reliability is high,
A unit section in which no error is detected by the error detection means is reproduced based on the input speech coding parameter, and a unit section in which an error is detected by the error detection means is selected by the selection means. A speech decoding apparatus comprising: reproduction means for reproducing an audio signal based on the audio encoding parameter.

Images