JP3583550B2 - Interpolator - Google Patents

Description

[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention relates to an interpolation device that calculates an output audio signal by performing parameter interpolation on a frame-by-frame basis for an audio encoded signal to which an error has been added in a transmission path.
[0002]
[Prior art]
2. Description of the Related Art Conventionally, an interpolation apparatus that interpolates a speech coded signal to which an error has been added in a transmission path by using a parameter of a past frame has been disclosed in “Channel Coding For Digital Speech Transmission In The Japan Digital Cellular System, J.L.M. Those described in the Institute of Electronics, Information and Communication Engineers, Research Group on Wireless Communication Systems, RCS90-27) are known.
[0003]
FIG. 3 is a block diagram showing a configuration example of a conventional interpolation device. 3, in this example, a data unpacking unit 1, a feature parameter interpolation unit 2, a past feature parameter storage unit 3, and a speech decoding unit 4 are provided. 1a is a received signal, 1b is a feature parameter, 1c is an error determination signal, 1d is a past feature parameter, 1e is a correction feature parameter, and 1j is a decoded speech signal.
[0004]
The operation of the interpolation device configured as described above will be described below. First, the data unpacking unit 1 separates the received signal 1a into a characteristic parameter 1b and an error determination signal 1c, and transmits them to the characteristic parameter interpolating unit 2. The feature parameter interpolation unit 2 calculates a corrected feature parameter 1e. When it is determined that the feature parameter 1b is erroneous by the error determination signal 1c transmitted to the feature parameter interpolation unit 2, the past feature parameter 1d stored in the past feature parameter storage unit 3 is corrected. The signal is output to the audio decoding unit 7 as 1e. If the feature parameter 1b is determined not to be erroneous by the error determination signal 1c transmitted to the feature parameter interpolating unit 2, the feature parameter 1b is output to the speech decoding unit 7 as the corrected feature parameter 1e. Further, the correction feature parameter 1e is stored in the past parameter interpolation unit 3 for use in processing of the next frame. The audio decoding unit 7 calculates a decoded audio signal 1j from the correction feature parameter 1e.
[0005]
[Problems to be solved by the invention]
However, in the above-described conventional interpolation device, since it is determined whether or not the feature parameters are collectively incorrect, even if some non-erroneous feature parameters partially exist, the past feature parameters are not used. Therefore, there is a problem that a portion where a feature parameter is not erroneous is not effectively used.
[0006]
The present invention solves the above-described conventional problem. Even when a part of a feature parameter of speech is erroneous due to a transmission path error, a part of the feature parameter that is not erroneous is effectively used, and a decoded speech signal is It is an object of the present invention to provide an excellent interpolating device capable of reducing the distortion of the signal.
[0007]
[Means for Solving the Problems]
In order to achieve the above object, the interpolation device of the present invention determines whether the voice of the frame currently being decoded is voiced or unvoiced using a correction feature parameter indicating a voice feature in a voiced / unvoiced determination unit, By switching the interpolation method in the power interpolation unit according to the determination result, even if the amplitude information of the unvoiced part is wrong, the amplitude of the unvoiced part is not decoded using the amplitude of the voiced part, so that the distortion in the decoded voice signal is reduced. This does not occur, and the distortion of the decoded speech signal can be reduced even when some of the elements in the speech feature parameters are erroneous due to transmission path errors.
[0008]
BEST MODE FOR CARRYING OUT THE INVENTION
The invention according to claim 1 of the present invention is directed to an interpolation apparatus for interpolating parameters in units of frames when a speech coded signal is erroneous due to a transmission path error. Data unpacking means for separating into a power parameter, which is the amplitude parameter of the audio signal, and a power parameter which is the amplitude information of the audio signal, and an error determination signal for individually determining whether or not the power parameter is incorrect. A feature parameter interpolating means for calculating a corrected feature parameter from a feature parameter output from the data unpacking means, an error determination signal, and a past feature parameter output from the past feature parameter storing means. The corrected feature parameter output by the feature parameter interpolation means. Voiced unvoiced determination means for determining whether the frame is voiced or unvoiced, past power parameter storage means for storing power parameters of a past frame, power parameters output from the data unpacking means, and the voiced unvoiced determination. Power correction means for determining the power of the decoded speech signal from the voiced / unvoiced flag calculated by the means and the past power parameters output by the past power parameter storage means; a correction feature parameter output from the parameter interpolation means; and the power Audio decoding means for calculating a decoded audio signal using the correction power parameter calculated by the correction means, wherein even if some of the audio characteristic parameters are erroneous due to a transmission path error, the characteristic parameters To make good use of the parts that are not mistaken, especially for some of the elements in the speech feature parameters. Even if the amplitude information of the sound that is wrong by the transmission path errors, such an action can be reduced distortion of the decoded speech signal.
[0010]
Hereinafter, embodiments of the present invention will be described with reference to the drawings.
(Embodiment 1)
FIG. 1 is a block diagram showing the configuration of the interpolation device according to the first embodiment of the present invention. In FIG. 1, 1 is a data unpacking unit that separates an input signal sequence into a plurality of signals, 2 is a feature parameter interpolation unit that interpolates an encoded audio signal as needed, and 3 is an encoded audio signal of a past frame. A past feature parameter storage unit 4 to be stored is a voiced / unvoiced determination unit that determines whether the voice of the frame currently being decoded is voiced or unvoiced, 5 is a power interpolation unit that interpolates amplitude information as necessary, and 6 is a past interpolation unit. Is a past power storage unit for storing the amplitude information of the frame, and 7 is an audio decoding unit for decoding audio.
[0011]
1a is a received signal, 1b is a feature parameter, 1c is an error determination signal, 1d is a past feature parameter, 1e is a voiced and unvoiced flag, 1g is a power parameter, 1h is a past power parameter, and 1i is a corrected power parameter. The power parameter 1j is the decoded audio signal.
[0012]
The operation of the interpolation device configured as described above will be described below. First, the data unpacking unit 1 separates the received signal 1a into a characteristic parameter 1b, an error determination signal 1c, and a power parameter 1g. As shown in FIG. 4, the feature parameter 1b is a set of a plurality of elements, and the error determination signal 1c is a set of signals indicating whether each element in the feature parameter 1b is erroneous.
[0013]
The feature parameter interpolation unit 2 calculates a corrected feature parameter 1e. At this time, the element in the feature parameter 1b determined to be erroneous by the error determination signal 1c transmitted to the feature parameter interpolating unit 2 is the element of the past feature parameter 1d stored in the past feature parameter storage unit 3. Replaced by each corresponding element. Each element of the feature parameter 1b determined not to be erroneous by the error determination signal 1c is output as it is as the corrected feature parameter 1e. Further, the correction feature parameter 1e is stored in the past parameter interpolation unit 3 for use in processing of the next frame.
[0014]
The voiced / unvoiced determination unit 4 determines whether the audio signal of the frame currently being decoded is voiced or unvoiced, and calculates a voiced / unvoiced flag 1f. The determination is made by using the correction feature parameter 1e. For example, if the speech pitch or similar information is included in the correction feature parameter 1e, whether or not the pitch of the frame currently being decoded is stationary, and the frequency response If there is information, the method is used to determine whether the frequency of the frame currently being decoded is close to noise.
[0015]
The power interpolation unit 5 performs interpolation of the power parameter 1g. When it is determined in the error determination signal 1c that the power parameter 1g is not incorrect, the power parameter 1g is output as the interpolation power parameter 1i. When it is determined in the error determination signal 1c that the power parameter 1g is erroneous and the voiced / unvoiced flag 1f indicates voiced, the power interpolation unit 5 corrects the past power parameter 1h stored in the past power storage unit 6 to correct the power. Output as parameter 1i. When it is determined in the error determination signal 1c that the power parameter 1g is erroneous and the voiced / unvoiced flag 1f indicates unvoiced, the power interpolation unit 5 outputs an amplitude value very close to no sound as the corrected power parameter 1i. Further, the corrected power parameter 1i is stored in the past power storage unit 6 for use in processing of the next frame.
[0016]
The audio encoding unit 7 calculates a decoded audio signal 1j using the correction feature parameter 1e and the correction power parameter 1i.
[0017]
By performing the above-described processing, a part that is not erroneous in the feature parameter is effectively utilized, and even when the amplitude information of the unvoiced part of the audio signal is lost due to a transmission path error, the unvoiced part is amplified and decoded. By not doing so, it is possible to obtain a decoded audio signal whose waveform information is not distorted.
[0018]
(Embodiment 2)
FIG. 2 is a block diagram showing the configuration of an interpolation device according to the second embodiment of the present invention. In FIG. 2, 1 is a data unpacking unit that separates an input signal sequence into a plurality of signals, 2 is a feature parameter interpolation unit that interpolates an encoded audio signal as needed, and 3 is an encoded audio signal of a past frame. A past feature parameter storage unit for storing, 7 a speech decoding unit for decoding speech, 8 a long-term prediction delay interpolation unit for interpolating a long-term prediction delay, 9 a past long-term prediction delay for storing a long-term prediction delay of a past frame It is a storage unit.
[0019]
1a is a received signal, 1b is a feature parameter, 1c is an error determination signal, 1d is a past feature parameter, 1e is a decoded speech signal, 1j is a decoded speech signal, 1k is a long-term prediction delay parameter, and 1l is a past long-term prediction delay. The parameter group 1m is a corrected long-term prediction delay parameter.
[0020]
The operation of the interpolation device configured as described above will be described below. First, the data unpacking unit 1 separates the received signal 1a into a characteristic parameter 1b, an error determination signal 1c, and a long-term prediction delay parameter 1k. As shown in FIG. 4, the feature parameter 1b is a set of a plurality of elements, and the error determination signal 1c is a set of signals indicating whether each element in the feature parameter 1b is erroneous.
[0021]
The feature parameter interpolation unit 2 calculates a corrected feature parameter 1e. At this time, the element in the feature parameter 1b determined to be erroneous by the error determination signal 1c transmitted to the feature parameter interpolating unit 2 is the element of the past feature parameter 1d stored in the past feature parameter storage unit 3. Replaced by each corresponding element. Each element of the feature parameter 1b determined not to be erroneous by the error determination signal 1c is output as it is as the corrected feature parameter 1e. Further, the correction feature parameter 1e is stored in the past parameter interpolation unit 3 for use in processing of the next frame.
[0022]
The long-term prediction delay interpolation unit 8 performs interpolation of a long-term prediction delay 1k. If it is determined in the error determination signal 1c that the long-term prediction delay 1k is not incorrect, the long-term prediction delay parameter 1k is output as the interpolated long-term prediction delay parameter 1m. When it is determined that the long-term prediction delay parameter 1k is incorrect in the error determination signal 1c, the long-term prediction delay interpolation unit 8 performs the plurality of past long-term prediction delay parameter groups 11 stored in the past long-term prediction delay storage unit 9. , And one numerical value in the long-term prediction delay parameter group 11 closest to the average value is output as the corrected long-term prediction delay parameter 1m. The corrected long-term prediction delay parameter 1m is stored in the past long-term prediction delay storage unit 9 for use in processing of the next frame, and discards the oldest data in the past long-term prediction delay parameter group.
[0023]
The speech encoding unit 7 calculates a decoded speech signal 1j using the corrected feature parameter 1e and the corrected long-term prediction delay parameter 1m.
[0024]
By performing the above processing, it is possible to make effective use of the part that is not erroneous in the feature parameter, and not to interpolate with half pitch or double pitch data even if the long-term prediction delay parameter is lost due to a transmission path error. As a result, it is possible to obtain a decoded audio signal whose waveform information is not distorted.
[0025]
【The invention's effect】
As described above, the present invention makes effective use of a part that is not erroneous even if the characteristic parameter of the audio is partially incorrect due to the transmission path, and particularly, when the amplitude information of the audio is incorrect, the characteristic parameter of the audio is used. By judging whether it is voiced or unvoiced, and by providing a means for switching the interpolation method of the amplitude information according to the result, even if the amplitude information of the unvoiced portion of the audio signal is lost due to a transmission path error, the unvoiced portion is amplified. By not decoding, a decoded audio signal whose waveform information is not distorted can be obtained.
[Brief description of the drawings]
FIG. 1 is a block diagram showing a configuration of an interpolation device according to a first embodiment of the present invention; FIG. 2 is a block diagram showing a configuration of an interpolation device according to a second embodiment of the present invention; FIG. FIG. 4 is a block diagram showing a configuration of an interpolation device. FIG. 4 is a conceptual diagram of feature parameters and error determination information.
REFERENCE SIGNS LIST 1 data unpacking unit 2 feature parameter interpolation unit 3 past feature parameter storage unit 4 voiced / unvoiced determination unit 5 power interpolation unit 6 past power storage unit 7 speech decoding unit 8 long-term prediction delay interpolation unit 9 past long-term prediction delay storage unit 1a reception Signal 1b Feature parameter 1c Error determination signal 1d Past feature parameter 1e Correction feature parameter 1f Voiced unvoiced flag 1g Power parameter 1h Past power parameter 1i Correction power parameter 1j Decoded speech signal 1k Long-term prediction delay parameter 1l Past long-term prediction delay parameter group 1m Long-term forecast delay parameter

Claims (1)

In an interpolator for interpolating parameters on a frame basis when a speech coded signal is erroneous due to a transmission path error, a received signal sent from a transmission path is a characteristic parameter indicating a characteristic of a speech signal and amplitude information of the speech signal. Data unpacking means for separating a power parameter and a feature parameter and an error determination signal for individually determining whether or not the power parameter is incorrect; past feature parameter storage means for storing feature parameters of a past frame; A feature parameter interpolating means for calculating a corrected feature parameter from the feature parameter output from the unpacking means, an error determination signal, and a past feature parameter output from the past feature parameter storage means, and a corrected feature parameter output from the feature parameter interpolating means Whether the frame is voiced or unvoiced Voiced / unvoiced determining means, power parameter of a past frame, a power parameter output from the data unpacking means, and a voiced / unvoiced flag calculated by the voiced / unvoiced determining means. A power correction unit that determines the power of the decoded audio signal from the past power parameters output by the past power parameter storage unit; and a correction feature parameter output from the parameter interpolation unit and a correction power parameter calculated by the power correction unit. An audio decoding means for calculating a decoded audio signal by using the interpolation apparatus.

Images