JPH07202821A - High efficiency coding voice decoder - Google Patents

Abstract

PURPOSE:To eliminate the need for addition of hardware such as a distribute filter and a synthesis filter after complete decoding by providing various adjustment to a high efficiency coding voice signal on the way of voice decoding. CONSTITUTION:A de-format section 12 analyzes a bit stream subject to high efficiency coding to separate allocation information, a scale factor and a frequency sampling thereby implementing de-formatting and the result is fed to a reproduction section 14, in which inverse quantization of a frequency sampling and inverse scaling processing are implemented and its output is given to an inverse mapping section 16 and a decoded acoustic signal is outputted. The scale factor of the reproduction section 14 is fed to an output section 23, from which a maximum value for each prescribed period of the scale factor is outputted as range information. Furthermore, a quantization coefficient in the reproduction section 14 is optionally adjusted by an input section 21.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a high-efficiency coded speech decoding apparatus for decoding a high-efficiency coded speech signal.

[0002]

2. Description of the Related Art Conventionally used speech compression is ADPCM utilizing correlation on a time axis, quasi-instantaneous compression, etc. The high-efficiency coded speech signal referred to here is a spectrum analysis of an acoustic signal. According to the human psychology model of human beings, a large amount of compression is performed by reducing the information about inaudible components.

As shown in FIG. 2 (A), a signal and a bit stream which are highly efficient coded are divided into a header and an audio signal into frequency bands, and scale factors and
It is composed of allocation information (allocation information) indicating the information amount of the frequency sample, a scale factor indicating the range (index information) of the band, and the frequency sample.

FIG. 11A shows a conventional high efficiency speech decoding apparatus.

The input bit stream 111 is allocated to the allocation information 11 in the deformat 112.
3a, the scale factor 113b, and the frequency sample 113c are divided with reference to the allocation information 113a. Divided frequency samples 113c
Is inversely quantized and inversely scaled by the reproducing unit 114. The inverse scaled frequency samples 115 are
The inverse mapping unit 116, which performs a frequency-time axis conversion operation such as IFFT or IMDCT, converts the data into time axis data, performs processing such as overlap, and then synthesizes the data to restore the sound signal 117.

The restored acoustic signal 117 is sent to the distribution filter 118. The distribution filter 118 frequency-decomposes the sound and divides it into a plurality of bands, and the level adjusting unit 120
To transmit. The input unit 121 outputs information 122 for giving the increase / decrease ratio of each band.
As a result, the level adjusting unit 120 performs amplification and attenuation for each band, and the output thereof is transmitted to the integrated filter 126. Further, the adjusted signal 125 of each band is determined to have a maximum value for a certain period, for example, 100 ms, and is sent to the output unit 123 for each band. Output unit 123
Then, the level of each band is displayed by an indicator or the like. The integrated filter 126 divides the signal 12 into frequency components.
5 is synthesized and output as an acoustic signal 127.

Next, the processing in the reproducing section 114 will be described with reference to FIG.

The dequantization unit 134 converts the reformatted frequency samples 113c into allocation information 113.
Based on a, it is inversely quantized and transmitted to the inverse scaling unit 136. Inverse quantization is performed by the following formula.

S '= c * (s + d) Here, c and d are called inverse quantization coefficients and are determined by the number of quantization steps indicated by the allocation information. Table 1 shows the inverse quantized coefficients c and d.

The coefficient calculator 137 obtains a multiplier 138 required for inverse scaling from the scale factor 113b, which is range information, by referring to a table or by calculation, and transmits a multiplier value 138 required for inverse scaling.
When the scale factor 113b represents 6 bits and a discrete value of 2 dB steps, the multiplier value 138 required for inverse scaling is calculated in dB, that is, from the value of the exponent, using a table or the like to calculate the multiplier value. I won't. As an example, Table 2 shows the relationship between the scale factor 113b and the multiplier value. The inverse scaling unit 136 multiplies the inversely quantized signal by the multiplier value 138 of each band obtained by the coefficient calculator 137, and outputs the result.

[0011]

According to the above-described highly efficient speech coded signal decoding apparatus, when an application for manipulating sound quality such as an equalizer is performed, complete decoding is performed once as shown in FIG. After the acoustic signal 117 has been obtained, the distribution filter distributes each band for adjustment and range display, and finally the adjusted signals are combined. Such hardware is required, which is disadvantageous in terms of circuit scale and price.

Therefore, the present invention eliminates the need for hardware such as a distribution filter and an integrated filter, allows various voice adjustments in the decoding process, and is a highly efficient coded voice decoding device capable of extracting range information. The purpose is to provide.

[0013]

[Means for Solving the Problems]

(1) Analyze the high efficiency coded bit stream,
Means for separating and reformatting allocation information, scale factor, frequency samples, means for outputting the maximum value or average value of the scale factor or frequency data for each fixed period, and the maximum value or average value The gist of the present invention is to have means for displaying, means for dequantizing frequency samples by the scale factor, and means for frequency integrating the dequantized frequency samples.

(2) A means for analyzing a bit stream and separating and reformatting allocation information, scale factor, and frequency samples, means for inputting values for operating frequency samples, and scale factor or frequency depending on the data of the input means. The gist of the present invention is to have means for changing data, means for dequantizing frequency samples by the scale factor, and means for frequency integrating the dequantized frequency samples.

[0015]

Operation: Before the bit stream is analyzed, the allocation information, the scale factor, the frequency samples are separated and reformatted, and the dequantized frequency samples are frequency integrated, the scale factor or the frequency data is kept constant. By outputting the maximum value or average value for each period and displaying it, and by inputting the value for operating the frequency sample, and changing the scale factor or frequency data, the analysis factor and integrated filter are unnecessary. There is.

[0016]

Embodiments of the present invention will be described below with reference to the drawings.

FIGS. 1A and 1B are block diagrams showing the overall construction of an embodiment of the present invention.

As shown in FIG. 2 (A), the signal (bit stream) coded with high efficiency is obtained by dividing the header and the audio signal into frequency bands, and indicates the scale factor and the information amount of frequency samples. It is composed of allocation information (allocation information), a scale factor indicating the range of the band (index information), and frequency samples.

FIG. 1A shows the first embodiment of the high-efficiency decoding device of the present invention.

The input bit stream 11 is
In the reformatting unit 12, allocation information 1
3a, the scale factor 13b, and the frequency samples divided by referring to the allocation information 13a are sent to the reproducing unit 14.

The reproducing unit 14 quantizes and descales the divided frequency sample 13c, and transmits it to the inverse mapping unit 16. The inverse mapping unit 16 performs frequency integration. Further, the maximum value or the average value of each band of the scale factor 13b or the inversely scaled frequency sample 15 for a certain fixed period, for example, 100 ms, is obtained and sent to the output unit 23. Output unit 23
Then, the level of each band is displayed by an indicator (not shown) which is a display means.

Further, the inversely scaled frequency sample 15 is subjected to IFFT (Inverse Fourier Transform) or IMDCT.
An inverse mapping unit 16 that performs a frequency-time axis conversion operation such as (inverse discrete cosine transform) frequency-integrates and converts the data into time-axis data, further performs processing such as overlap and synthesizes the acoustic signal. Restored as 17.

The first embodiment is configured as described above, and the reproducing section 14 uses a scale factor or frequency data, as will be described later, to input a signal to the reproducing section 14 or a signal output from the reproducing section 14. The range information of can be obtained. That is, unlike the conventional technique, it is not necessary to provide a means for obtaining range information after complete decoding.

FIG. 1B shows a second embodiment of the high efficiency decoding apparatus of the present invention.

The input bit stream 11 is divided by the deformatting unit 12 into allocation information 13a, scale factor 13b and frequency samples 13c with reference to the allocation information 13a and sent to the reproducing unit 14.

The input section 21 inputs the increase / decrease rate of each band and sends the control signal 22 to the reproducing section 14. The reproduction unit 14 amplifies each band by the control signal 22,
The frequency sample 13 c that has been attenuated and divided is quantized, inversely scaled, and transmitted to the inverse mapping unit 6.

Descaled frequency samples 15
Is converted into time axis data by an inverse mapping unit 16 such as IFFT or IMDCT, which performs frequency-time axis conversion calculation, is subjected to processing such as overlap, is synthesized, and is restored as an acoustic signal 17.

The second embodiment is configured as described above, and the reproducing unit 14 can obtain various audio adjustments by using the control signal 22 from the input unit 21 as described later. That is, unlike the conventional case, it is not necessary to provide a means for performing voice adjustment after complete decoding.

Next, various examples of the reproducing section 14 realized in FIGS. 1A and 1B will be described.

FIG. 2B shows processing blocks of the reproducing section 14 of FIG. 1A.

The dequantization section 34 dequantizes the frequency-formatted frequency sample 13c based on the allocation information 13a, dequantizes it, and transmits it to the descaling section 36. Inverse quantization is performed by the following formula.

S '= c * (s + d) Here, c and d are called inverse quantization coefficients and are determined by the number of quantization steps indicated by the allocation information. An example of the inverse quantized coefficients c and d is shown in FIG.

The coefficient calculator 37 obtains a multiplier 38 required for inverse scaling from the scale factor 13b, which is range information, by referring to a table or by calculation, and transmits the required multiplier value 38 to the inverse scaling unit 36. When the scale factor 13b represents 6 bits and a discrete value of 2 dB steps, the multiplier value 38 required for inverse scaling must be calculated from dB, that is, the value of the exponent, using a table or the like. . As an example thereof, FIG. 4 shows the relationship between the scale factor 13b and the multiplier value. In the inverse scaling unit 36, the inversely quantized signal is multiplied by the multiplier value 38 of each band obtained by the coefficient calculator 37.
Is multiplied and output.

(Function as Indicator) FIG. 2
In (B), the scale factor is also input to the comparator / adder 40. The comparator / adder 40 compares the scale factors for a certain fixed period, for example, 5 frames (1 frame is 20 ms), obtains the maximum value, and outputs it as the range information 24 of each band in the period.

Alternatively, the comparator / adder 40 has a fixed period of time, for example, 5 frames (one frame is 20 ms).
Add the scale factors of and average (total value)
Is obtained and is output as the range information 24 of each band in the period.

The above example has been described as an example of obtaining the range information 24 of the input signal input to the reproducing unit 14. However, the range information may be obtained from the output signal of the reproducing unit 14.

As shown in FIG. 2C, the output of the inverse scaling section 36 is supplied to the comparator / adder 40. The comparator / adder 40 performs square addition of the inversely scaled frequency samples 15 of, for example, 5 frames to obtain an average value (total value), and the range information 2 of each band in the period.
Output as 4.

(Function as Equalizer) FIG.
The reproduction unit 14 of FIG. 1B is shown.

The dequantization section 34 dequantizes the frequency-formatted sample 13c based on the allocation information 13a, and transmits it to the descaling section 36. Inverse quantization is performed by the following formula.

S '= c * (s + d) Here, c and d are called inverse quantization coefficients and are determined by the number of quantization steps indicated by the allocation information. The inverse quantization coefficients c and d are shown in FIG.

The coefficient calculator 37 calculates the multiplier 3 required for inverse scaling from the scale factor 13b which is range information.
8 is obtained by referring to a table or by calculation, and a multiplier value 38 required for inverse scaling is transmitted. When the scale factor 13b represents 6 bits and a discrete value of 2 dB steps, the multiplier value 38 required for inverse scaling is calculated in dB from the exponent value by using a table or the like. I won't. As an example thereof, FIG. 4 shows the relationship between the scale factor 13b and the multiplier value.

Here, the scale factor 13b or the multiplier value 38 is increased or decreased for each band according to the input value 22. Also, since the scale factor is a discrete value, when it is changed to the next value by the input value 22, it changes abruptly. Therefore, instead of changing the scale factor value, it is possible to smoothly change it by applying a window function to the multiplier value ( 6A), the generation of noise is reduced.

The inverse scaling unit 36 multiplies the inversely quantized signal by the multiplier value 38 of each band obtained by the coefficient calculator 37, and outputs it.

Further, as shown in FIG. 6B, a level change value 42 for each band is determined only by having an adjustment value table 41 for the multiplier change and selecting various modes, and this is calculated by the coefficient calculator 37. You may give it to.

As described above, various applications can be realized by changing the scale factor value or the multiplier value of each band.

Next, various kinds of applications to which the equalizer function as described above is applied will be described.

(LR balance) In the configuration of FIG. 1B, when adjusting the input section 21, it is possible to group not only the values for each band but also several bands and change them simultaneously.

FIG. 7 shows an example of the input panel of the input section 21. For example, when changing the scale factors of all the bands at the same time for each channel, if the L channel is used, the total adjustment 81 of only the L channel may be increased or decreased, or the LR adjustment 82 may be increased or decreased to lower the L channel while changing the R channel. An application that adjusts the balance of the LR by raising or the like is possible.

(Bass Boost) Input unit 21 of FIG. 1 (B)
If you send the information that selected the bass boost mode to
As the adjustment value table 41 in (B), the table shown in FIG. 8A is applied to correct the scale factor.
As a result, the effect emphasizing the reduction is realized. That is, data is set in the adjustment value table so that a gain that emphasizes the low frequency range can be obtained according to the frequency band.

(Loudness Correction) Human auditory sense has different sensitivity depending on the frequency of the sound, and when the sound pressure is changed, the change in the auditory sense differs depending on the band. For example, when you turn down the volume, the high range and low range feel more quiet than the mid range. It is easy to correct it.

The value of the volume (sound pressure level adjustment) of the output acoustic signal or the information for selecting the loudness mode is sent to the input section 21 of FIG. 1 (B). At this time, when the volume is low or the loudness mode is selected, the adjustment value table 41 shown in FIG. 6 (B) sends a table as shown in FIG. It is set to increase the multiplier value in the high range. Alternatively, an appropriate table is selected or calculated by calculation according to the volume amount.

(Sound Effect; Frequency Changer, Tone Change) Frequency samples are exchanged between bands (see FIG. 9A) or copied to another band to artificially create overtones (FIG. 9B). )), Or by shifting the band of the frequency samples to realize pitch shift (see FIG. 9C), and other various special effects can be realized. This is because the quantized signal is a quantized coefficient divided in frequency bands, so that it is easy to change the coefficient or to add each coefficient.

Further, the above operation is performed before the inverse quantizer 34,
That is, it may be performed at the stage of the frequency sample 13c or after the inverse quantization, that is, at the stage of the signal 35.

(Echo function) As shown in FIG. 10A, the frequency sample 101 and the feedback 103
The echo effect can be obtained by adding and with the adder 102 to form the output signal 104. The output signal 104 is sent to the buffer 106, and the buffer 106 delays its input by an externally set delay amount and outputs it. The output of the buffer 106 is input to the attenuator 105, attenuated by an externally set attenuation rate, and sent to the adder 102. As described above, it is possible to perform echo for each band by setting different delay amounts and attenuation rates for each band.

The above echo device is the same as the signal 1 of the reproducing section of FIG.
3c, signal 35, or signal 15.

(Various Hall Reproducing Functions) In the echo device shown in FIG. 10 (A), different delay amounts for each band,
It is also possible to prepare some representative values (samples) using the attenuation rate as a table and select the table by external selection input.

In this way, the data (table) of various concert halls and the like are prepared and the state of various concert halls and the like is realized only by selecting.

(Wrap Function) In FIG. 10A, the audio data buffer 10 for a certain fixed period, for example, 0.3 seconds.
6, the data captured in the buffer 106 is repeatedly output, for example, three times, and the output signal is the attenuator 1
It is possible to realize the wrap function by outputting the signal 103 as it is as the signal 104 without passing the frequency sample in the adder 102 through 05. That is, the adder 102 is used as a selector. After that (0.9 seconds in this case)
The adder 102 outputs the frequency sample 101 as the signal 104 without adding the signal 103. The wrap effect can be achieved by repeating the operation of fetching the 0.3 second data in the buffer 106. This period and the number of times can be set arbitrarily. Also, even if it does not always take a constant value, it recognizes the passage of time, the state of voice data, such as when the sound pressure has decreased, as a phrase,
You may change the number of times.

FIG. 10B shows the timing chart.

Since the reproducing unit is usually processed by a DSP (digital signal processor) or the like, the above various applications can be realized only by changing the software.

[0061]

As described above, according to the present invention, various applications can be executed without the addition of distribution filters and synthesis filters.

[Brief description of drawings]

FIG. 1 is a block diagram showing an embodiment of the present invention.

FIG. 2 is a diagram showing a bit stream of high-efficiency coded audio and a block diagram showing details of a reproducing unit in FIG.

FIG. 3 is a diagram showing an example of inverse quantization coefficients.

FIG. 4 is a diagram showing an example of a scale factor.

FIG. 5 is a block diagram showing details of a reproducing unit in FIG.

FIG. 6 is a diagram shown for explaining a state when a scale factor is changed and a diagram showing another example of a reproducing unit.

FIG. 7 is a diagram showing an example of an input panel.

FIG. 8 is a diagram shown for explaining an example of a sound adjusting function of a reproducing unit.

FIG. 9 is a diagram shown for explaining an example of another audio adjusting function of the reproducing unit.

FIG. 10 is a diagram shown for explaining an example of still another audio adjusting function of the reproducing unit.

FIG. 11 is a diagram showing a conventional high-efficiency coded speech signal decoding apparatus.

[Explanation of symbols]

12 ... Deformat, 14 ... Reproducing unit, 16 ... Inverse mapping unit, 21 ... Input unit, 23 ... Output unit, 34 ... Inverse quantization unit, 36 ... Inverse scaling unit, 37 ... Coefficient calculator, 40
... comparator / adder.

Claims (6)

[Claims] 1. Deformatting means for separating and reformatting allocation information, scale factor, and frequency samples from a high-efficiency coded bit stream, and range information for outputting the maximum value of the scale factor for each constant period. Output means, display means for displaying the maximum value, dequantization means for dequantizing the frequency samples according to the scale factor, frequency integration of frequency samples dequantized by the dequantization means And a frequency unifying means for performing high efficiency coding speech decoding apparatus. 2. The means for outputting the maximum value, further comprising means for outputting the maximum value of the dequantized frequency samples for each constant period instead of the scale factor. High-efficiency coded speech decoding device. 3. The decoding apparatus for high efficiency coded speech according to claim 1, further comprising means for outputting an average value instead of the maximum value in the means for outputting the maximum value. 4. The decoding apparatus for high efficiency coded speech according to claim 2, further comprising means for outputting an average value instead of the maximum value in the means for outputting the maximum value. 5. Deformatting means for separating and reformatting allocation information, scale factor, and frequency samples from a high-efficiency coded bit stream, and input means for inputting values for operating the frequency samples. Adjusting means for changing the scale factor according to the data of the input means, and the scale factor obtained from the adjusting means,
An apparatus for decoding highly efficient coded speech, comprising: an inverse quantizer for inversely quantizing the frequency sample; and a means for frequency integrating the frequency samples inversely quantized by the inverse quantizer. 6. Decoding of high-efficiency coded speech according to claim 5, further comprising means for changing the dequantized frequency sample instead of the scale factor in the adjusting means for changing the scale factor. Device.