JP3147127B2 - Speech waveform coding device - Google Patents

Abstract

PURPOSE:To obtain the speech waveform encoding device which prevents a high-frequency noise from being generated by suppressing abrupt variation in the voiceless output of the output waveform of a speech decoding means at a border point. CONSTITUTION:This device is equipped with a predicting means 14 which calculates a predicted value, a subtracting means 11 which subtracts the predicted value from an input PCM code to calculate a difference PCM code, an adaptive quantizing means 12 which converts the difference PCM code into a quantized value, an adaptive inverse quantizing means 13 which requantizes the quantized value and outputs the requantized value, an adding means 15 which adds the requantized value and the predicted value together, a minimum code output means 17 that outputs a quantized value which corresponds to a quantization width coefficient less than 1 and has the same sign with the last requantized value in terms of time and corresponds to a value less in numeral than it, and a selecting means 18 which selects the quantized value from the adaptive quantizing means 12 or the quantized value from the minimum code output means 17 according to a voiceless output indication and outputs it to the decoding means.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a speech waveform coding apparatus, and more particularly to a speech waveform coding apparatus capable of forming a code for forcibly outputting silence.

[0002]

2. Description of the Related Art In recent years, cordless telephones using analog transmission have become widespread, but problems such as limitations on channel frequency allocation and voice leakage have arisen. Therefore, with the development of digital signal processing theory and the advancement of IC miniaturization technology,
Cordless telephones using digital transmission have been developed. Among them, in order to reduce the amount of data on a transmission path, a speech codec that performs speech encoding and decoding has been developed. ADPCM, a waveform encoding method that is easy to process,
(Adaptive difference PCM) is about to be adopted.

[0003] One use of a digital cordless telephone is to use it outdoors. The ambient noise is much higher outdoors than indoors. Therefore, when used outdoors, the content from the partner cannot be heard correctly, and the partner is uncomfortable. Therefore, when the user is not talking due to large ambient noise, the encoded voice is not output, but a silent code is output instead. As a result, the other party does not feel uncomfortable. This is the silent output function.

FIGS. 12 and 13 show a conventional example.

[0005] The encoding circuit 50 in FIG. 12 includes a prediction circuit 14 for calculating a predicted value, a switch 56 that is turned on / off in response to a silent output instruction, an input buffer 57 that stores a differential PCM code through the switch 56, and an input buffer 57. A selection circuit 58 for selecting either the output PCM code or the output of the input buffer 57 in accordance with the silent output instruction signal, and subtracting the output of the selection circuit 58 and the output of the prediction circuit 14 to calculate a differential PCM code Subtraction circuit 11, an adaptive quantization circuit 12 that converts the differential PCM code calculated by the subtraction circuit 11 into a quantized value, and outputs the quantized value to a decoding circuit via a transmission path. An adaptive inverse quantization circuit 13 that inversely quantizes the quantized value, an addition circuit 15 that adds the output of the adaptive inverse quantization circuit 13 and the output of the prediction circuit 14 and outputs the result to the prediction circuit 14. It is configured.

Here, as shown in FIG. 2, the adaptive quantization circuit 12 includes a quantization width buffer 125 for storing the quantization width calculated from the past ADPCM code, and a subtraction circuit 1
A second multiplying circuit 124 for multiplying the differential PCM code from 1 by the quantization width from the quantization width buffer 125;
Quantization / coefficient table 1 which quantizes the output of the multiplication circuit 124 and outputs the quantization value and the quantization width coefficient to the adaptive inverse quantization circuit 13 or the decoding circuit 60 via the transmission path.
21 is multiplied by the output of the quantization width buffer 125 and the output of the quantization / coefficient table 121 to obtain the quantization width buffer 12
5 and a first multiplication circuit 123 for outputting the result to the first multiplication circuit 123.

The adaptive inverse quantization circuit 13 includes an inverse quantization table 131 for converting the ADPCM code from the selection circuit 18 into an inverse quantization value, as shown in FIG. 3, and an ADPCM code from the adaptive selection circuit 18. Is converted into a quantization width coefficient, and a quantization width buffer 135 that stores the quantization width calculated from the past ADPCM code.
And output of coefficient table 132 and quantization width buffer 13
A third multiplication circuit 133 that multiplies the output of the quantization width buffer 135 by the output of the third multiplication circuit 135 and outputs the output of the quantization width buffer 135 to the addition circuit 15. And a multiplication circuit 134.

Next, the operation of the above conventional example will be described.

Here, the PCM code input at time t is “1”, the output of the prediction circuit 14 is “3”, and the quantization width buffer 125 and the quantization width buffer 135 are calculated from the past ADPCM code. It is assumed that the obtained quantization width “1” is stored.

The quantization / coefficient table 121 has an ADP when the quantization value is "+3" as shown in FIG.
“011” (binary number) as the CM code, “0.8” as the quantization width coefficient, and ADP when the quantization value is “+2”
“010” (binary number) as the CM code, “1” as the quantization width coefficient, and ADPCM when the quantization value is “+1”
“001” (binary number) as the code, “1.25” as the quantization width coefficient, and ADPCM when the quantization value is “0”
When the code is “000” (binary number), the quantization width coefficient is “1.25”, and when the quantization value is “−1”, ADPC
“100” (binary number) as the M code, “1.25” as the quantization width coefficient, and ADP when the quantization value is “−2”
"101" (binary number) as the CM code, "1" as the quantization width coefficient, and ADPCM when the quantization value is "-3".
It is assumed that “110” (binary number) is described as a code and “0.8” is described as a quantization width coefficient.

As shown in FIG. 7, when the ADPCM code is "011" (binary number), "+3" is shown as the inverse quantization value, and the ADPCM code is shown in the inverse quantization table 131 and the inverse quantization table 631. When the value is “010” (binary number), “+2” is used as the inverse quantization value. When the ADPCM code is “001” (binary number), “+” is used as the inverse quantization value.
When the ADPCM code is "000" (binary number), "0" is used as the inverse quantization value, and when the ADPCM code is "10".
When it is "0" (binary number), "-1" is used as an inverse quantization value,
When the ADPCM code is “101” (binary number), the inverse quantization value is “−2”, and the ADPCM code is “110”.
In the case of (binary number), "-3" is described as the inverse quantization value.

Coefficient table 132 and coefficient table 6
32, the ADPCM code is "0" as shown in FIG.
11 "(binary number)," 1.2
When the ADPCM code is “010” (binary number), “1” is used as the quantization width coefficient, and the ADPCM code is “0”.
01 "(binary number)," 0.
When the ADPCM code is “000” (binary number), “0.8” is used as the quantization width coefficient. When the ADPCM code is “100” (binary number), the quantization width coefficient is “0. 8 ”is ADPCM code“ 101 ”(binary number)
In this case, it is assumed that "1" is described as the quantization width coefficient, and that when the ADPCM code is "110" (binary), "1.25" is described as the quantization width coefficient.

First, the ADPCM encoding when there is no silent output instruction at time t will be described.

Since there is no silent output instruction, the switch 56 is turned on, and the selection circuit 58 selects the input PCM code. Then, the PCM code is transferred to the input buffer 57 and the subtraction circuit 11. The subtraction circuit 11 outputs a difference PCM code “−2” obtained by subtracting the output “3” of the prediction circuit 14 from the input PCM code “1”. Next, the second multiplier 124 in the adaptive quantization circuit 12 calculates the difference PC
M code “−2” and contents of quantization width buffer 125 “1”
And the output “−2” is quantized / coefficient table 1
Transfer to 21. In the quantization / coefficient table 121, the second
Of the multiplication circuit 124 is compared with each of the quantized values, and the closest quantized value is selected.
The CM code “101” (binary number) and the quantization width coefficient “1” are output.

The ADPCM code "101" (binary number) is transferred to the adaptive inverse quantization circuit 13 and, at the same time, is transferred to the decoding circuit through a transmission path. Further, the quantization width coefficient “1” and the quantization width “1” stored in the quantization width buffer 125 are multiplied by the first multiplier 123, and the output “1” is supplied to the quantization width buffer 125. Will be returned. The ADPCM code “101” (binary number) is converted into an inverse quantization value “−2” and a quantization width coefficient “1” by an inverse quantization table 131 and a coefficient table 132 in the adaptive inverse quantization circuit 13, respectively. You.

The output “−2” of the inverse quantization table 131 and the quantization width “1” stored in the quantization width buffer 135 are multiplied by a fourth multiplier 134. The quantization width coefficient “1” is multiplied by the quantization width “1” stored in the quantization width buffer 135 by the third multiplier 133, and the output “1” is output from the quantization width buffer 135.
Is returned to. Next, the output “−2” of the fourth multiplication circuit 134
Is added to the output of the prediction circuit 14 by the addition circuit 15 as a difference value. The result of the addition is returned to the prediction circuit 14 to calculate the next predicted value.

Next, a case where a silent output is instructed at time t + 1 will be described.

The switch 56 is turned off by a silence output instruction signal instructing to output silence, and the selection circuit 5
8 selects the output of the input buffer 57. That is, the PCM code “1” immediately before the silent output instruction, that is, the PCM code “1” at the time t is stored in the input buffer 57, and the PCM code is transferred to the subtraction circuit 11 through the selection circuit 58. The subtraction circuit 11 subtracts the output “1” of the prediction circuit 14 from the PCM code “1” from the input buffer 57 and a difference PCM code “0”.
Is output.

Next, the second multiplier 124 in the adaptive quantization circuit 12 multiplies the difference PCM code "0" by the content "1" of the quantization width buffer 125, and quantizes the output "0". Transfer to the coefficient table 121. The quantization / coefficient table 121 compares the output “0” of the second multiplication circuit 124 with each quantization value, selects the closest quantization value, and outputs the corresponding ADPCM code “000” (binary number). The quantization width coefficient “1.25” is output. ADPCM code "00
"0" (binary number) is transferred to the decoding circuit through the transmission path at the same time as being transferred to the adaptive inverse quantization circuit 13.

The quantization width coefficient “1.25” is multiplied by the quantization width “1” stored in the quantization width buffer 125 by the first multiplication circuit 123, and the output “1.2” is output.
"5" is returned to the quantization width buffer 125. ADPCM
The code “000” (binary number) is the inverse quantization value “0” and the quantization width coefficient “0.8” in the inverse quantization table 131 and the coefficient table 132 in the adaptive inverse quantization circuit 13, respectively.
Is converted to The fourth multiplier 134 multiplies the output “0” of the inverse quantization table 131 by the quantization width “1” stored in the quantization width buffer 125. The quantization width coefficient “0.8” is stored in the quantization width buffer 135.
And the third multiplication circuit 13
The result is multiplied by 3, and the output “0.8” is returned to the quantization width buffer 135. Next, the output “0” of the fourth multiplication circuit 134 is used as a difference value by the prediction circuit 14 in the addition circuit 15.
Is added to the output of The result of the addition is obtained by the prediction circuit 14 again.
To calculate the next predicted value.

The above processing is repeated until the silent output instruction signal is cut off.

As shown in FIG. 13, the output waveform from the decoding circuit when the sample after time t + 1 is instructed to output silence is such that a DC waveform outside the audible range is output and becomes silent. The waveform shown by the dotted line in FIG. 13 is a waveform assuming that no instruction is given.

[0023]

However, in the above-mentioned conventional example, the waveform output from the speech decoding circuit has a disadvantage that a sudden change occurs at a boundary point of silence output, so that high-frequency noise is generated. Was.

[0024]

SUMMARY OF THE INVENTION An object of the present invention is to improve the disadvantages of the prior art, and in particular, to prevent the waveform output from the speech decoding circuit from changing abruptly at the boundary of the silent output, thereby reducing the generation of high-frequency noise. It is an object of the present invention to provide a speech waveform encoding device that can prevent such a problem.

[0025]

Accordingly, the present invention provides a prediction circuit for calculating a prediction value, a subtraction circuit for subtracting the output of the prediction circuit from an input PCM code and calculating a differential PCM code, and a subtraction circuit. An adaptive quantizer for converting the differential PCM code calculated in the above into a quantized value, an adaptive inverse quantizer for inversely quantizing the quantized value from the adaptive quantizer and outputting the inverse quantized value, An addition circuit for adding the output of the quantization circuit and the output of the prediction circuit, and corresponding to a quantization width coefficient smaller than 1 and having the same value as the sign of the immediately preceding inverse quantization value and a numerical value less than that A minimum code output circuit for outputting a quantized value corresponding to the value, and an adaptive inverse quantization circuit and a decoding circuit for selecting one of the output of the adaptive quantization circuit and the output of the minimum code output circuit in accordance with the silent output instruction signal And a selection circuit for outputting to It adopts a configuration in that. This aims to achieve the above-mentioned object.

[0026]

In the case where there is no silent output instruction: the subtraction circuit subtracts the output of the prediction circuit from the input PCM code in the subtraction circuit.
Output sign. Next, in the adaptive quantization circuit, the difference PC
Multiplying the M code by a pre-stored quantization width,
A quantization value and a quantization width coefficient are output based on the result of the multiplication. In the adaptive quantization circuit, the quantization width coefficient is multiplied by a previously stored quantization width, and the multiplication result is stored as a new quantization width. Further, the selection circuit selects a quantization value output from the adaptive quantization circuit, and this quantization value is transferred to the adaptive inverse quantization circuit,
The data is transferred to the decoding circuit through the transmission path. In the adaptive inverse quantization circuit, the quantization value is converted into an inverse quantization value and a quantization width coefficient. The inverse quantization value is multiplied by a quantization width stored in advance, and the multiplication result is transferred to an addition circuit and a minimum sign output circuit. The quantization width coefficient is multiplied by a previously stored quantization width, and the result of the multiplication is stored as a new quantization width. In the adder circuit, the output of the adaptive inverse quantization circuit is added to the output of the prediction circuit, and the result of the addition is returned to the prediction circuit, and the prediction circuit calculates the next predicted value. Further, in the minimum code output circuit, the output of the adaptive inverse quantization circuit is converted into a quantization value.

When a silent output instruction is issued: The output of the minimum code output circuit is selected in the selection circuit by the silent output instruction signal. That is, the output of the minimum code output circuit is transferred to the adaptive inverse quantization circuit and the transmission path through the selection circuit.
In the adaptive inverse quantization circuit, the quantization value from the minimum code output circuit is converted into an inverse quantization value and a quantization width coefficient, and the inverse quantization value is multiplied by a quantization width stored in advance. The data is transferred to the minimum code output circuit and is converted into a quantized value. Further, in the adaptive inverse quantization circuit, the quantization width coefficient is multiplied by a previously stored quantization width, and the multiplication result is stored as a new quantization width. The above processing is repeated until the silence output instruction signal is cut off.

[0028]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS One embodiment of the present invention will be described below with reference to FIGS.

The encoding circuit 1 in the first embodiment shown in FIG.
0 is the prediction circuit 14 for calculating the prediction value and the input P
A subtraction circuit 11 for subtracting the prediction value from the prediction circuit 14 from the CM code to calculate a differential PCM code; an adaptive quantization circuit 12 for converting the differential PCM code calculated by the subtraction circuit 11 into a quantization value; Adaptive inverse quantization circuit 1 that inversely quantizes the quantization value from quantization circuit 12 and outputs the inverse quantization value
3, an addition circuit 15 for adding the inverse quantization value from the adaptive inverse quantization circuit 13 and the prediction value from the prediction circuit 14, and an inverse quantization coefficient corresponding to a quantization width coefficient smaller than 1 and immediately preceding in time. A minimum code output circuit 17 for outputting a quantized value corresponding to a value which is the same as the sign of the quantized value but smaller than the numerical value, and a quantized value from the adaptive quantizing circuit 12 in response to a silent output instruction. It comprises an adaptive inverse quantization circuit 13 which selects one of the quantized values from the code output circuit 17 and outputs it to a decoding circuit 60 via a transmission line.

Here, the adaptive quantization circuit 12 includes a quantization width buffer 125 for storing the quantization width calculated from the past ADPCM code as shown in FIG.
A second multiplying circuit 124 for multiplying the differential PCM code from 1 by the quantization width from the quantization width buffer 125;
Quantization / coefficient table that quantizes the output of the multiplication circuit 124 and outputs the quantized value and the quantization width coefficient to the adaptive inverse quantization circuit 13 or the adaptive inverse quantization circuit 63 of the decoding circuit 60 via the transmission path. 121 and the quantization width buffer 12
5 multiplied by the output of the quantization / coefficient table 121 and output to the quantization width buffer 125
23.

The adaptive inverse quantization circuit 13 includes an inverse quantization table 131 for converting the quantization value from the adaptive quantization circuit 12 into an inverse quantization value, as shown in FIG. , A coefficient table 132 for converting the quantized value from into a quantization width coefficient, a quantization width buffer 135 for storing the quantization width calculated from the past ADPCM code, an output of the coefficient table 132 and the quantization width buffer 135 , And a third multiplier 133 for multiplying the output of the quantization width buffer 135 by the output of the inverse quantization table 131 and the output of the quantization width buffer 135. Multiplying circuit 1 for outputting to
34.

Further, the decoding circuit 60 has a prediction circuit 64 for calculating a prediction value as shown in FIG. 4 and an adaptive circuit for inversely quantizing the quantization value sent from the encoding circuit 10 via the transmission line. It comprises an inverse quantization circuit 63 and an addition circuit 65 that adds the output of the adaptive inverse quantization circuit 63 to the output of the prediction circuit 64, outputs the result to the prediction circuit 64, and outputs it as a decoded PCM code.

Here, the adaptive inverse quantization circuit 63 includes an inverse quantization table 631 for converting the quantization value from the encoding circuit 10 into an inverse quantization value as shown in FIG. , A coefficient table 632 for converting the quantization value into a quantization width coefficient, a quantization width buffer 635 for storing the quantization width calculated from the past ADPCM code, and an output of the coefficient table 632 and the quantization width buffer 635. A third multiplication circuit 633 that multiplies the output by the output of the quantization width buffer 635 and a fourth multiplication by which the output of the inverse quantization table 631 and the output of the quantization width buffer 635 are multiplied and output to the addition circuit 65 And a circuit 634.

Next, the operation of the first embodiment will be described.

The PCM code input at time t is "1",
The output of the prediction circuit 14 and the prediction circuit 64 is set to “3”,
Quantization width buffer 125 and quantization width buffer 135
, The quantization width “1” calculated from the past ADPCM code is stored.

In the quantization / coefficient table 121, when the quantization value is "+3" as shown in FIG.
“011” (binary number) as the code, “0.8” as the quantization width coefficient, and ADPCM when the quantization value is “+2”
The code is “010” (binary number), the quantization width coefficient is “1”. When the quantization value is “+1”, the ADPCM code is “001” (binary number), and the quantization width coefficient is “1.25”. Is ADPCM when the quantization value is “0”.
When the code is “000” (binary number), the quantization width coefficient is “1.25”, and when the quantization value is “−1”, ADPC
“100” (binary number) as the M code, “1.25” as the quantization width coefficient, and ADP when the quantization value is “−2”
"101" (binary number) as the CM code, "1" as the quantization width coefficient, and ADPCM when the quantization value is "-3".
It is assumed that “110” (binary number) is described as a code and “0.8” is described as a quantization width coefficient.

As shown in FIG. 7, when the ADPCM code is “011” (binary number), “+3” is shown as the inverse quantization value, and the ADPCM code is shown in the inverse quantization table 131 and the inverse quantization table 631. When the value is “010” (binary number), “+2” is used as the inverse quantization value. When the ADPCM code is “001” (binary number), “+” is used as the inverse quantization value.
When the ADPCM code is "000" (binary number), "0" is used as the inverse quantization value, and when the ADPCM code is "10".
When it is "0" (binary number), "-1" is used as an inverse quantization value,
When the ADPCM code is “101” (binary number), the inverse quantization value is “−2”, and the ADPCM code is “110”.
In the case of (binary number), "-3" is described as the inverse quantization value.

Coefficient table 132 and coefficient table 6
32, the ADPCM code is "0" as shown in FIG.
11 "(binary number)," 1.2
When the ADPCM code is “010” (binary number), “1” is used as the quantization width coefficient, and the ADPCM code is “0”.
01 "(binary number)," 0.
When the ADPCM code is “000” (binary number), “0.8” is used as the quantization width coefficient. When the ADPCM code is “100” (binary number), the quantization width coefficient is “0. 8 ”is ADPCM code“ 101 ”(binary number)
In this case, it is assumed that "1" is described as the quantization width coefficient, and that when the ADPCM code is "110" (binary), "1.25" is described as the quantization width coefficient.

First, ADPCM encoding / decoding in the case where there is no silent output instruction at time t will be described. The subtraction circuit 11 outputs a difference PCM code “−2” obtained by subtracting the output “3” of the prediction circuit 14 from the input PCM code “1”. Next, the second multiplication circuit 124 in the adaptive quantization circuit 12 multiplies the differential PCM code “−2” by the content “1” of the quantization width buffer 125, and outputs the output “−2” by quantization / coefficient. Transfer to table 121. Quantization / coefficient table 121
Compares the output “−2” of the second multiplication circuit 124 with each quantized value, selects the nearest quantized value “−2”, and, as shown in FIG. 6, the corresponding ADPCM code “10”.
1 (binary number) and a quantization width coefficient "1" are output. Since there is no silent output instruction, the selection circuit 18
2 selects the ADPCM code output by
The DPCM code “101” (binary number) is transferred to the decoding circuit 60 through the transmission path at the same time as being transferred to the adaptive inverse quantization circuit 13.

The quantization width coefficient “1” is multiplied by the quantization width “1” stored in the quantization width buffer 125 by the first multiplying circuit 123, and the output “1” is obtained by the quantization width “1”. It is returned to the buffer 125. The ADPCM code "10
“1” (binary number) is an inverse quantization table 131 and a coefficient table 132 in the adaptive inverse quantization circuit 13, respectively.
Are converted into the inverse quantization value "-2" and the quantization width coefficient "1". The fourth multiplier 134 multiplies the output “−2” of the inverse quantization table 131 by the quantization width “1” stored in the quantization width buffer 135. The quantization width coefficient “1” is stored in the quantization width buffer 13.
5 is multiplied by the quantization width “1” stored in the third multiplication circuit 133, and the output “1” is returned to the quantization width buffer 135.

Next, the output "-"
"2" is transferred to the addition circuit 15 and the minimum code output circuit 17 as a difference value. Prediction circuit 14 in addition circuit 15
Is added to the output of The result of the addition is obtained by the prediction circuit 14 again.
To calculate the next predicted value. The minimum sign output circuit 17 inputs the difference value “−” as shown in FIG.
Since the quantization width coefficient corresponding to “2” is “1”, the ADPCM code “100” (2
Base number). On the other hand, the ADPCM code “101” (binary number) sent to the decoding circuit 60 through the transmission line is inversely quantized by the inverse quantization table 631 and the coefficient table 632 in the adaptive inverse quantization circuit 63 as shown in FIG. It is converted into a quantization value “−2” and a quantization width coefficient “1”.

The output “−2” of the inverse quantization table 631 and the quantization width “1” stored in the quantization width buffer 635 are multiplied by a fourth multiplier 634. The quantization width coefficient “1” is different from the quantization width “1” stored in the quantization width buffer 635 by the third multiplier 633.
, And the output “1” is supplied to the quantization width buffer 63.
Returned to 5. Next, the output “−−” of the fourth multiplication circuit 634
"2" is added to the output of the prediction circuit 64 by the addition circuit 65 as a difference value. The result of the addition is output as a PCM code. It is returned to the prediction circuit 64 for the next sample.

Next, a case where a silent output is instructed at time t + 1 will be described. In response to a silence output instruction signal for instructing to output silence, the selection circuit 18 causes the minimum code output circuit 1
7 is selected. That is, the output “100” (binary number) of the minimum code output circuit 17 is transferred to the adaptive inverse quantization circuit 13 and the transmission path through the selection circuit 18. That AD
PCM code "100" (binary number) is an adaptive inverse quantization circuit 1
3, the inverse quantization table 131 and the coefficient table 13
2, the inverse quantization value "-1" as shown in FIG.
And a quantization width coefficient “0.8”. Output “−1” of the inverse quantization table 131 and the quantization width buffer 13
The fourth multiplication circuit 134 multiplies the quantization width “1” stored in No. 5 by “4”.

The quantization width coefficient “0.8” is different from the quantization width “1” stored in the quantization width buffer 135 by the third.
, And the output “0.8” is returned to the quantization width buffer 135. Next, the output “−1” of the fourth multiplication circuit 134 is transferred to the minimum sign output circuit 17 and converted into “100”. On the other hand, the ADPCM code “100” sent to the decoding circuit 60 through the transmission path
(Binary) is converted into an inverse quantization value “−1” and a quantization width coefficient “0.8” in the inverse quantization table 631 and the coefficient table 632 in the adaptive inverse quantization circuit 63, respectively, as shown in FIG. Is converted. The fourth multiplying circuit 634 multiplies the output “−1” of the inverse quantization table 631 by the quantization width “1” stored in the quantization width buffer 635. The quantization width coefficient “0.8” is multiplied by the quantization width “1” stored in the quantization width buffer 635 by the third multiplier 633, and the output “0.8” is quantized. Returned to width buffer 635.

Next, the output of the fourth multiplying circuit 634 "-
"1" is added to the output of the prediction circuit 64 by the addition circuit 65 as a difference value. The result of the addition is output as a PCM code. Also, it is transferred to the prediction circuit 64 for the next sample. The above processing is repeated until the silence output instruction signal is cut off.

As shown in FIG. 8, the output waveform from the decoding circuit 60 when the sample after the time t + 1 is instructed to output a silent sound is a DC waveform outside the audible range and becomes silent. Note that the decimal places are rounded off here.
The waveform shown by the dotted line in FIG. 8 is a waveform assuming that no instruction is given.

The encoding circuit 3 in the second embodiment shown in FIG.
0 is the prediction circuit 14 for calculating the prediction value and the input P
The output of the prediction circuit 14 is subtracted from the CM code, and the difference PCM
A subtraction circuit 11 for calculating a code, and an adaptive quantization circuit 32 for converting the differential PCM code calculated by the subtraction circuit 11 into a quantization value and a quantization width coefficient in accordance with a silent output instruction signal and outputting the converted value to a decoding circuit 40 An adaptive inverse quantization circuit 13 that inversely quantizes the quantized value output from the adaptive quantization circuit 32; an addition that adds the output of the adaptive inverse quantization circuit 13 and the output of the prediction circuit 14 and outputs the result to the prediction circuit 14 And a circuit 15.

Here, the adaptive quantization circuit 32 stores a quantization width buffer 1 for storing the quantization width calculated immediately before in time.
25, a switch 321 that is turned on / off in response to a silence output instruction, a silence buffer 322 that stores a differential PCM code through the switch 321, and a difference PCM code and an output of the silence buffer 322 according to a silence output instruction signal. A first selection circuit 323 that performs
Multiplying the output of the second multiplier 124 by the output of the quantization width buffer 125, quantizing the output of the second multiplier 124, transferring the quantized value to the adaptive inverse quantization circuit 13, and simultaneously transmitting the And a quantization / coefficient table circuit 121 for outputting a quantization width coefficient to the second selection circuit 325 and a coefficient output circuit 3 for outputting a quantization width coefficient smaller than 1 which is predetermined.
24, the quantization width coefficient output by the quantization / coefficient table circuit 121 in response to the silent output instruction signal, and the coefficient output circuit 3
24, and a second selection circuit 325 that selects one of the outputs of the second selection circuit 24 and transfers it to the decoding circuit 40 through the transmission path. And a first multiplication circuit 123 for calculating

Here, the adaptive inverse quantization circuit 13 has the same configuration and function as the first embodiment.

Further, the decoding circuit 40 includes a prediction circuit 64 for calculating a predicted value as shown in FIG. 10 and an adaptive inverse quantization for inversely quantizing the quantized value of the ADPCM code transmitted via the transmission path. And a summation circuit 65 that adds the output of the adaptive inverse quantization circuit 43 to the output of the prediction circuit 64 and outputs the result as a decoded PCM code.

Here, the adaptive inverse quantization circuit 43 stores an inverse quantization table 431 for converting the ADPCM code from the encoding circuit 30 into an inverse quantization value, and a quantization width calculated from the past ADPCM code. Quantization width buffer 435
A third multiplication circuit 433 that multiplies the quantization width coefficient from the encoding circuit 30 by the output of the quantization width buffer 435 and outputs the result to the quantization width buffer 435;
And a fourth multiplying circuit 434 that multiplies the output of the 31 and the output of the quantization width buffer 435 and outputs the result to the adding circuit 65.

Next, the operation of the second embodiment will be described.

The PCM code input at time t is “1”,
The outputs of the prediction circuit 14 and the prediction circuit 44 are set to “3”,
Quantization width buffer 125 and quantization width buffer 435
, The quantization width “1” calculated from the past ADPCM code is stored.

It is assumed that the inverse quantization table 431 describes a table similar to the inverse quantization table 131 in the first embodiment.

First, ADPCM encoding / decoding in the case where there is no silent output instruction at time t will be described. The subtraction circuit 11 outputs a difference PCM code “−2” obtained by subtracting the output “3” of the prediction circuit 14 from the input PCM code “1”. Since there is no silent output instruction, the switch 321 in the adaptive quantization circuit 32 is turned on, and the first selection circuit 323 determines the difference P
Select the CM code. Thereby, the differential PCM code is transferred to the silence buffer 322 and the second multiplying circuit 124.

The second multiplier 124 multiplies the differential PCM code “−2” by the content “1” of the quantization width buffer 125 and outputs the output “−2” to the quantization / coefficient table 121.
Transfer to The quantization / coefficient table 121 compares the output “−2” of the second multiplication circuit 124 with each quantized value, selects the nearest quantized value “−2”, and as shown in FIG. Corresponding ADPCM code "101" (binary number)
And the quantization width coefficient “1”.

ADPCM output from adaptive quantization circuit 32
The code “101” (binary number) is transferred to the adaptive inverse quantization circuit 13 and simultaneously to the decoding circuit 40 via the transmission path. Further, since there is no silent output instruction, the second selection circuit 325 sets the quantization / coefficient table 1 as shown in FIG.
21 selects the quantization width coefficient “1” output. As a result, the quantization width coefficient “1” is transferred to the first multiplication circuit 123 and, at the same time, is transmitted to the decoding circuit 40 through the transmission path.
Transferred to The quantization width coefficient “1” is multiplied by the quantization width “1” stored in the quantization width buffer 125 by the first multiplication circuit 123, and the output “1” is supplied to the quantization width buffer 125. Will be returned. The ADPCM code "1
"01" (binary number) is converted into an inverse quantization value "-2" and a quantization width coefficient "1" in the inverse quantization table 131 and the coefficient table 132 in the adaptive inverse quantization circuit 13, respectively, as shown in FIG. Is converted. The fourth multiplier 134 multiplies the output “−2” of the inverse quantization table 131 by the quantization width “1” stored in the quantization width buffer 135. The quantization width coefficient “1” is stored in the quantization width buffer 13.
5 is multiplied by the quantization width “1” stored in the third multiplication circuit 133, and the output “1” is returned to the quantization width buffer 135.

Next, the output "-"
"2" is transferred to the addition circuit 15 as a difference value. 4th
The output “−2” of the multiplication circuit 134 is added to the output of the prediction circuit 14 by the addition circuit 15. The result of the addition is returned to the prediction circuit 14 to calculate the next predicted value. On the other hand, the ADPCM code “101” (binary number) sent to the decoding circuit 40 through the transmission path is converted into an inverse quantization value “−2” by the inverse quantization table 431 in the adaptive inverse quantization circuit 43. The multiplication circuit 434 multiplies the output “−2” of the inverse quantization table 431 by the quantization width “1” stored in the quantization width buffer 435. The quantization width coefficient “1” sent to the decoding circuit 40 through the transmission path is multiplied by the quantization width “1” stored in the quantization width buffer 435 by the multiplication circuit 433, and the output “ "1" is returned to the quantization width buffer 435. Next, the output “−2” of the multiplication circuit 434 is added to the output of the prediction circuit 44 by the addition circuit 45 as a difference value. The result of the addition is output as a PCM code. It is returned to the prediction circuit 44 for the next sample.

Next, a case where a silent output is instructed at time t + 1 will be described. The switch 321 is turned off by the silence output instruction signal instructing to output silence, and the first
Selection circuit 323 selects the output of the silence buffer 322. That is, the output “−2” of the silence buffer 322 is transferred to the second multiplier 124 through the first selector 323. Next, the second multiplication circuit 124 calculates the difference P
The CM code “−2” is multiplied by the content “1” of the quantization width buffer, and the output “−2” is quantized / coefficient table 121.
Transfer to The quantization / coefficient table 121 compares the output “−2” of the second multiplication circuit 124 with each quantized value, selects the nearest quantized value “−2”, and as shown in FIG. Corresponding ADPCM code "101" (binary number)
And the quantization width coefficient “1”. Adaptive quantization circuit 32
Are output to the adaptive inverse quantization circuit 13 and at the same time are transferred to the decoding circuit 40 through the transmission path.

In addition, the second selection circuit 325 selects the quantization width coefficient “0.6” output from the silent coefficient output circuit 324 according to the silent output instruction signal. Accordingly, the quantization width coefficient “0.6” is transferred to the decoding circuit 40 through the transmission path at the same time as being transferred to the first multiplication circuit 123.
In the first multiplying circuit 123, the quantization width coefficient “0.6” is converted to the quantization width “1” stored in the quantization width buffer 125.
, And the output “0.6” is input to the quantization width buffer 1
It is returned to 25. On the other hand, the decoding circuit 40
The ADPCM code “101” (binary number) sent to is converted to an inverse quantization value “−2” by an inverse quantization table 431 in the adaptive inverse quantization circuit 43. The multiplication circuit 434 multiplies the output “−2” of the inverse quantization table 431 by the quantization width “1” stored in the quantization width buffer 435. Further, the quantization width coefficient “0.6” sent to the decoding circuit 40 through the transmission path is multiplied by the quantization width “1” stored in the quantization width buffer 435 by the third multiplication circuit 433. , Its output “0.6” is returned to the quantization width buffer 435. Next, the output “−1” of the fourth multiplication circuit 434 is added to the output of the prediction circuit 44 by the addition circuit 45 as a difference value. The result of the addition is output as a PCM code. Also, it is transferred to the prediction circuit 44 for the next sample. The above processing is repeated until the silence output instruction signal expires.

The output waveform from the decoding circuit 40 when the sample after time t + 1 is instructed to output a silent signal is shown in FIG.
As shown in (2), a DC waveform outside the audible range is output, resulting in no sound. Note that the decimal places are rounded off here. The waveform shown by the dotted line in FIG. 11 is a waveform assuming that no instruction is given.

In the first embodiment, the ADPCM encoding / decoding in the backward system has been described. In the second embodiment, the ADPCM encoding / decoding in the forward system has been described. If a circuit for changing the quantization width is used, the same effect as the present invention can be obtained.

When the adaptive difference code is lost, the first
In the embodiment, the minimum code output circuit 17 outputs the second ADPCM code such that the absolute value of the inverse quantization value is minimized.
In the embodiment described above, the inverse quantization value outputs the value immediately before the disappearance, and the output circuit for silence outputs a fixed value of “0.6”. However, the inverse quantization value is the sign of the value immediately before the disappearance. Is the same, the absolute value is less than 0 except for 0, and if the quantization width coefficient is smaller than “1”, the PCM code gradually becomes a constant value. The effect is obtained.

In the first embodiment, the minimum code output circuit 17 and the selection circuit 18 required for the present invention are located at the subsequent stage of the quantization / coefficient table 121. In the second embodiment, the silence required for the present invention is provided. Buffer 322 and the first selection circuit 323 are located at the preceding stage of the quantization / coefficient table 121,
ADPC in the latter stage of the table including the coefficient table
If the M code value is used and the differential PCM code is used in the preceding stage, the same effects as those of the present invention can be obtained regardless of the coding method.

The present embodiment has been described using the quantization table. However, if the ADPCM code matches the data expression of the operation as it is, the same effects as those of the present invention can be obtained without the inverse quantization table. .

Although the present embodiment has been described using hardware blocks for easy understanding, the same effects as those of the present invention can be obtained even if the present invention is realized by software of a general computer. For example, an adaptive quantization circuit and an adaptive inverse quantization circuit use two multiplication circuits. However, if only one multiplication circuit or only an addition circuit is used in a time-division manner by software, it is the same as the present invention. Effects can be obtained.

[0067]

Since the present invention is constructed and functions as described above, according to the present invention, it is possible to prevent an abrupt change in the output waveform at the boundary point of silence output, thereby preventing high frequency noise from being generated. It is possible to provide an unprecedented excellent speech waveform encoding device that does not cause discomfort to the listener.

[Brief description of the drawings]

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

FIG. 2 is a block diagram illustrating an adaptive quantization circuit in the backward scheme shown in FIG. 1 in detail.

FIG. 3 is a block diagram illustrating an adaptive inverse quantization circuit in FIG. 1 in detail;

FIG. 4 is a block diagram showing a decoding circuit in the backward scheme shown in FIG. 1;

FIG. 5 is a block diagram illustrating an adaptive inverse quantization circuit in FIG. 4 in detail;

FIG. 6 is an explanatory diagram showing contents of a quantization / coefficient table in FIG. 2;

FIG. 7 is an explanatory diagram showing contents of an inverse quantization table and a coefficient table in FIG. 3;

FIG. 8 is an output waveform diagram in the first embodiment shown in FIG.

FIG. 9 is a block diagram showing a second embodiment of the present invention.

FIG. 10 is a block diagram of a decoding circuit in the forward system of FIG. 9;

FIG. 11 is an output waveform diagram in the second embodiment shown in FIG. 9;

FIG. 12 is a block diagram showing a conventional encoding circuit of a backward type to which a silence output function is added.

FIG. 13 is an output waveform diagram in the conventional example shown in FIG.

[Explanation of symbols]

 11: subtraction circuit 12: adaptive quantization circuit 13: adaptive inverse quantization circuit 14: prediction circuit 15: addition circuit 17: minimum code output circuit 18: selection circuit

──────────────────────────────────────────────────続 き Continued on front page (58) Field surveyed (Int.Cl. ⁷ , DB name) G10L 19/00

Claims (2)

(57) [Claims] 1. A prediction circuit for calculating a prediction value, and an output of the prediction circuit is subtracted from an input PCM code to obtain a difference P
A subtraction circuit for calculating a CM code; an adaptive quantization circuit for converting the differential PCM code calculated by the subtraction circuit into a quantization value; and an adaptive inverse quantization for inversely quantizing the quantization value from the adaptive quantization circuit. A quantization circuit, an addition circuit for adding the output of the adaptive inverse quantization circuit and the output of the prediction circuit, and a code corresponding to a quantization width coefficient smaller than 1 and having the same sign as the immediately preceding inverse quantization value. A minimum code output circuit that outputs a quantized value corresponding to a value whose numerical value is less than or equal to one of the output of the adaptive quantization circuit and the output of the minimum code output circuit in response to a silent output instruction signal. And an output circuit for selecting the adaptive inverse quantization circuit and the decoding circuit. 2. A prediction circuit for calculating a prediction value, and an output of the prediction circuit is subtracted from an input PCM code to obtain a difference P
A subtraction circuit for calculating a CM code, an adaptive quantization circuit for converting the differential PCM code calculated by the subtraction circuit into a quantization value and a quantization width coefficient according to a silence output instruction signal, and outputting the result to a decoding circuit; An adaptive inverse quantization circuit that inversely quantizes the quantization value output by the adaptive quantization circuit; and an addition circuit that adds the output of the adaptive inverse quantization circuit and the output of the prediction circuit and outputs the result to the prediction circuit. An audio waveform encoding device characterized by being constituted.