JP3227929B2 - Speech encoding apparatus and decoding apparatus for encoded signal - Google Patents

Abstract

PURPOSE:To provide the audio encoder capable of performing efficient encoding. CONSTITUTION:An audio signal SA is converted into a digital signal and is supplied to a discrimination circuit 13 so as to discriminate whether or not it is in a sound or in a silence part, then the discrimination signal SD is supplied to a time decision circuit 14. In the time decision circuit 14 obtains a timing signal ST corresponding to the sound part silence part of an audio signal SAD time-adjusted by a delay circuit 17. The timing signal ST is supplied to an encoding circuit 15 and compresses and encodes the sound part of the audio signal SAD. The time data DTa and DTb are encoded by an encoding circuit 16. The encoding signals obtained by encoding circuits 15 and 16 are supplied to a frame processing circuit 18 and are arranged in the proper format by associating encoding signals. As the silence part of the audio signal does not perform compression and encoding and encodes the only data of the continuous time, the effective encoding is enabled.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a speech encoding device and a decoding device for decoding the encoded signal.

[0002]

2. Description of the Related Art Normally, when an audio signal is encoded and transmitted, it is directly transmitted to an ADPCM irrespective of a sound part and a silent part.
(Adaptive Differential Pulse Code Modulation),
In many cases, compression coding such as SBC (Sub-Band Coding) and DCT (Discrete Cosine Transform) is performed.

[0003]

However, in general, there is a considerable proportion of silent portions in the audio signal, and it has been difficult to say that efficient encoding is performed.

In view of the above, the present invention provides a speech encoding device capable of performing efficient encoding and a decoding device for the encoded signal.

[0005]

According to a first aspect of the present invention, there is provided a speech encoding apparatus for discriminating whether an input speech signal is a sound part or a silent part, and the discrimination means.
The time of the sound part of the input audio signal is substantially set back and forth.
A voiced part extending means for time-expanding, and a part of the input audio signal which is determined to be a voiced part by the determination means is compression-encoded into a first part.
A first encoding means for obtaining a coded signal of the second type, and a second coding means for encoding data indicating a time of a portion of the input audio signal determined to be a silent part by at least the determination means to obtain a second coded signal
Encoding means.

[0006] audio decoding apparatus according to the second invention, sound
The part of the input audio signal that is determined to be
First decoding means for decoding the obtained first encoded signal
And at least a portion of the input audio signal determined to be a silent portion
The data indicating the time during the second encoded signal and the second decoding means that turn into decrypt, talkspurt of a speech signal obtained from the first decoding means obtained by encoding And a signal output unit that inserts a silent signal for a time corresponding to a silent section obtained by the decoding means.

[0007]

According to the first aspect of the present invention, only the sound portion of the input audio signal is compression-encoded, and the silent portion thereof is encoded only at the time without compression encoding. Efficient encoding can be performed. As a result, the recording (transmission) rate can be reduced. In addition, the sound part is extended back and forth.
Small amplitude part at the beginning and end of the part is missing
This can be prevented.

According to the second aspect of the present invention, a silent signal is inserted between audio signals of a sound part obtained by decoding for a time of a silent part obtained by decoding. It becomes possible to obtain good.

[0009]

An embodiment of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram illustrating a speech encoding device.

In FIG. 1, after an audio signal SA supplied to an input terminal 11 is converted into a digital signal by an A / D converter 12, the audio signal SA is a sound part (a part where an audio level exists). It is supplied to a sound / silence discrimination circuit 13 for judging whether or not the sound is a silent part (a part where no sound level exists). FIG. 2 shows a configuration example of the determination circuit 13.

In FIG. 2, an audio signal SA is supplied to an absolute value circuit 13a and converted into an absolute value. This absolute value circuit 1
The audio signal | SA | converted into an absolute value in 3a is supplied to the A input terminal of the comparator 13b, and is compared with a threshold value Vth supplied to the B input terminal. The comparator 13b outputs a high-level “H” signal when | SA | ≧ Vth, and outputs a low-level “L” signal when | SA | <Vth. Therefore, the comparator 1
The output signal of 3b is low level “L” when the audio signal SA crosses the 0 level in the sound part and in the silent part.
Becomes

The output signal of the comparator 13b is supplied to a series circuit of latch circuits 13c to 13e composed of D flip-flops, and the output signals of these latch circuits 13c to 13e are supplied to an OR circuit 13f. This allows
As the output signal of the OR circuit 13f, a signal in which the low level "L" at the portion crossing the 0 level in the sound portion is removed is obtained. The output signal of the OR circuit 13f is the latch circuit 1
The signal is output as a discrimination signal SD between a sound part and a silent part through 3g.

FIG. 3 shows signals of each part of the discrimination circuit 13. FIG. 7A shows an audio signal SA, and FIG.
b to the output signal of FIG.
13E show output signals of the OR circuit 13f, and FIG. G shows a determination signal SD output from the latch circuit 13g.

Returning to FIG. 1, the discrimination signal SD output from the discrimination circuit 13 is supplied to the time determination circuit 14. FIG.
Shows an example of the configuration of the time determination circuit 14, and the CPU 14
a and a clock circuit 14b.

The discrimination signal SD output from the discrimination circuit 13
Is supplied to the CPU 14a. The CPU 14a refers to the time data output from the clock circuit 14b to measure the rise time (start time of the sound part) and the fall time (end time of the sound part) of the determination signal SD. That is, when the audio signal SA is as shown in FIG. 5A,
Since the discrimination signal SD is obtained as shown in FIG. B, time t1 is the start time of the sound part, time t2 is the end time of the sound part,
Time t3 is measured as the start time of the sound part.

Then, in the CPU 14a, the audio signal SA _D (shown in FIG. 5C, DT
Is a high level "H" corresponding to a sound portion of the delay circuit (delay time of the delay circuit), and a timing signal ST (illustrated in FIG. D) is set to a low level "L" corresponding to a silent portion.
In this case, the sound part detected by the determination circuit 13 is extended back and forth by a predetermined time ΔT. That is, the time T of the sound part
a ′ is Ta + 2ΔT, and the silent time Tb ′ is Tb.
-2ΔT. The reason for extending the time of the sound part forward and backward by a predetermined time ΔT is that the discriminating circuit 13 discriminates the sound part by threshold processing, and the minute part of the start part and end part of the sound part. This is to prevent the amplitude portion from being lost.

The CPU 14a outputs a timing signal ST as an output signal of the time determination circuit 14, and sequentially outputs data DTa and DTb of the times Ta 'and Tb'.

Referring back to FIG. 1, the timing signal ST output from the time determination circuit 14 is
And the data DTa and DTb are supplied to the encoding circuit 16. The audio signal SA output from the A / D converter 12 is supplied to a delay circuit 17, and the audio signal SA _D delayed by DT in the delay circuit 17 (FIG. 5).
C) is supplied to the encoding circuit 15. Here, the delay time DT is calculated by the time determination circuit 14 as the time Ta ′, Tb ′.
Is set in consideration of the time required to determine the value and a predetermined time ΔT.

In the coding circuit 15, the sound part of the audio signal SA _D is compression-coded by a conventionally known coding method such as ADPCM, SBC, DCT or the like. The encoded signal of the sound part output from the encoding circuit 15 is supplied to a framing circuit 18.

In the encoding circuit 16, the time Ta ', Tb'
Are encoded.

FIG. 6 shows a configuration example of the encoding circuit 16. In the figure, reference numeral 161a denotes an average value circuit. Data DTa and DTb are supplied to the average value circuit 161a, and the average value D of the data DTa and DTb within a certain period is previously determined.
Taa and DTba are required. The average values DTaa and DTba output from the average value circuit 161a are
b to obtain quantized data QTaa and QTba.

Reference numeral 161c denotes a latch circuit. The latch circuit 161c stores data D at time Ta 'and Tb'.
Ta and DTb are supplied. A timing signal ST is supplied to the latch circuit 161c, and a latch operation is performed at the rising timing. As a result, the time T of the sound part where the compression coding is performed by the coding circuit 15 is calculated.
The data DTa of a 'and the data DTb of the time Tb' of the silent part following the sound part are latched.

The data DTa and DTb latched by the latch circuit 161c are supplied to a subtraction circuit 161d, and subtracted from the average values DTaa and DTab. Subtraction circuit 161
The difference values Δaa and Δba output from d are quantized by the quantization circuit 161.
e to obtain quantized data QΔaa and QΔba. Quantized data Q output from the quantization circuit 161b
Taa, QTba and the quantized data QΔaa, QΔba output from the quantization circuit 161e are supplied to the framing circuit 18 as coded signals.

FIG. 7 shows another example of the configuration of the encoding circuit 16. In the figure, 162a is a latch circuit,
Data DTa and DTb are supplied to the latch circuit 162a. In the latch circuit 162a, the encoding circuit 1
5, the time Ta 'of the first voiced part in which the compression encoding is performed.
Data DTa and the time T of the silent part following the sound part
The data DTb of b 'is latched as initial values DTa0 and DTb0. The initial values DTa0 and DTb0 output from the latch circuit 162a are supplied to a quantization circuit 162b, and quantized data QTa0 and QTb0 are obtained.

Reference numeral 162c denotes a latch circuit. The latch circuit 162c stores data D at time Ta 'and Tb'.
Ta and DTb are supplied. The timing signal ST is supplied to the latch circuit 162c, and the latch operation is performed at the rising timing. As a result, the time T of the sound part where the compression coding is performed by the coding circuit 15 is calculated.
The data DTa of a 'and the data DTb of the time Tb' of the silent part following the sound part are latched.

The data DTa and DTb latched by the latch circuit 162c are supplied to a subtraction circuit 162d, where the data DTa and DTb are subtracted from the initial values DTa0 and DTb0. Subtraction circuit 161
The difference values Δa0 and Δb0 output from d
e to obtain quantized data QΔa0, ΔQb0. Quantized data Q output from quantization circuit 162b
Ta0, QTb0 and the quantized data QΔa0, QΔb0 output from the quantization circuit 162e are supplied to the framing circuit 18 as encoded signals.

In the framing circuit 18, the encoding circuit 15
The compressed coded signal of the voiced part output from the multiplexing unit is associated with the coded signal of the data DTa and DTb of the voiced part and the silent part following the time Ta ′ and Tb ′ output from the coding circuit 16. And placed in the appropriate format.

When the encoding circuit 16 shown in FIG. 6 is used, the quantized data QTaa, QTba and QΔa are associated with an encoded signal of a sound part to be compressed and encoded first.
a, QΔba, and quantized data QΔaa, QΔ in association with a coded signal of a sound part that has been compression-coded from the second time on.
Only ba is arranged.

When the encoding circuit 16 shown in FIG. 7 is used, the quantized data QTa0 and QTb0 are arranged in association with the encoded signal of the sound part which has been compressed and encoded first, and the second and subsequent times. Quantized data QΔa0 and QΔb0 are arranged in association with a coded signal of a sound part compressed and coded.

The framed data derived from the framing circuit 18 to the output terminal 19 is subjected to error correction coding and channel coding (not shown), and is recorded or transmitted.

In the speech encoding apparatus shown in FIG. 1, the speech signal SA
Only the voiced part of _D is compression-coded, and the silent part thereof is not compression-coded and only the data of the duration is coded, so that efficient coding can be performed. Thereby, there is an advantage that the recording (transmission) rate can be reduced.

FIG. 8 is a block diagram showing a speech decoding apparatus for decoding an output signal of the speech encoding apparatus of FIG. 1 and obtaining an original speech signal. In the figure, framed data (output signal of the framing circuit 18 in FIG. 1) supplied to an input terminal 21 is supplied to a frame decomposition circuit 22.

The compressed coded signal of the sound part output from the frame decomposition circuit 22 is supplied to a decoding circuit 23 and decoded. Then, the audio signal of the sound part decoded by the decoding circuit 23 is sequentially written into the FIFO type memory 24. Here, when the audio signal of the sound part decoded by the decoding circuit 23 is output, a control signal is supplied from the decoding circuit 23 to the write address generation circuit 25, and the write address WAD is sequentially generated. The decoded audio signal of the sound part is written into the memory 24.

The coded signals of the data DTa and DTb of the time Ta 'and Tb' of the sound part and the subsequent silent part which are decoded by the decoding circuit 23 from the frame decomposition circuit 22 are supplied to the decoding circuit 26. The decoding circuit 26 obtains data DTa and DTb in the decoding process. Then, a control signal is supplied from the decoding circuit 26 to the read address generation circuit 27, and the read address RAD is sequentially generated during the time Ta ', and the voice signal of the sound part is read from the memory 24, and During the subsequent time Tb ', the generation of the read address RAD is stopped, and the state is substantially the same as the state where the silence signal is read from the memory 24.

The operation of writing the decoded audio signal of the sound part into the memory 24, reading the sound signal at the time Ta ', and stopping the reading during the subsequent time Tb'. Is repeated for each sound part. In this case, it is necessary to start writing at least the audio signal of the next sound part into the memory 24 before the time Tb 'ends.

On the output side of the memory 24, audio signals of a sound part and a silent part are obtained alternately and continuously. This memory 2
The output signal of No. 4 is converted into an analog signal by a D / A converter 28, and an audio signal SA before encoding is obtained at an output terminal 29.

In the example of FIG. 8, the reading of the memory 24 is stopped to reproduce the sound signal of the silent part. However, the silent signal is written in the memory 24 for the silent part time Tb '. Reading of the memory 24 may be performed continuously.

In the above embodiment, the time T of the silent section is
Although the data DTa of the voiced part time Ta 'is also encoded in addition to the data DTb of b', the audio decoding device does not need the data DTa unlike the example of FIG. In some cases, the encoding of this data DTa becomes unnecessary.

[0039]

According to the first aspect of the present invention, only the sound part of the input audio signal is compression-coded, and the silent part is not compressed and coded, but only the data of the duration thereof.
Efficient encoding can be performed, and the recording (transmission) rate can be reduced.

Further, by extending the sound part back and forth, it is possible to prevent the minute amplitude part at the start and end of the sound part from being lost. Further, by making the encoded signal of the time data of the sound part and the silent part into the quantized data of the difference value from the second time on, the encoding efficiency can be further improved.

According to the second aspect of the present invention, a silent signal is inserted between the audio signals of the sound part obtained by decoding for the time of the silent part obtained by decoding. Can be obtained favorably.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating a configuration of a speech encoding device according to an embodiment of the present invention.

FIG. 2 is a block diagram illustrating a configuration example of a determination circuit of the speech encoding device.

FIG. 3 is a diagram for explaining the operation of a discrimination circuit.

FIG. 4 is a block diagram illustrating a configuration example of a time determination circuit of the speech encoding device.

FIG. 5 is a diagram for explaining the operation of the time determination circuit.

FIG. 6 is a block diagram illustrating a configuration example of an encoding circuit of a speech encoding device.

FIG. 7 is a block diagram illustrating another configuration example of the encoding circuit of the speech encoding device.

FIG. 8 is a block diagram showing a configuration of a speech decoding device as an embodiment of the present invention.

[Description of Signs] 11, 21 Input terminal 13 Sound / silence discrimination circuit 14 Time determination circuit 15, 16 Encoding circuit 17 Delay circuit 18 Frame conversion circuit 19, 29 Output terminal 22 Frame decomposition circuit 23, 26 Decoding Circuit 24 Memory 25 Write address generation circuit 27 Read address generation circuit

Claims (4)

(57) [Claims] 1. A discriminating means for discriminating whether an input sound signal is a sound part or a soundless part, and a sound part of the input sound signal discriminated by the discriminating means .
Sound part expanding means for extending the time substantially back and forth for a predetermined time
And first encoding means for compressing and encoding a part of the input audio signal which is determined as a sound part by the determination means to obtain a first encoded signal; And a second encoding means for encoding data indicating the time of the portion of the input audio signal to obtain a second encoded signal. 2. The method according to claim 1, wherein the input audio signal is a sound part or a silent part.
Discriminating means for discriminating the presence or absence of the input voice signal,
A first code for compressing and encoding the portion to obtain a first coded signal;
Encoding means and the input which is determined to be a silent part by at least the determining means
The data indicating the time of the portion of the audio signal is encoded to form a second
Second encoding means for obtaining an encoded signal, wherein the second encoded signal has an average value of the time and
Data obtained by quantizing the difference between time and the average value
A speech coding apparatus characterized by the above-mentioned. 3. The method according to claim 1, wherein the input audio signal is a sound part or a silent part.
Discriminating means for discriminating the presence or absence of the input voice signal,
A first code for compressing and encoding the portion to obtain a first coded signal;
Encoding means and the input which is determined to be a silent part by at least the determining means
The data indicating the time of the portion of the audio signal is encoded to form a second
Second encoding means for obtaining an encoded signal, wherein the second encoded signal includes an initial value of the time and
Quantizes data indicating the difference between time and the initial value
A speech coding apparatus characterized by the above-mentioned. 4. A part of an input audio signal determined to be a sound part.
To decode a first encoded signal obtained by compression encoding
First decoding means, and at least a part of the input audio signal determined to be a silent part.
Encoding obtained by encoding data indicating the time of
A second decoding unit for decoding the signal; and a sound signal of the sound part obtained by the first decoding unit.
It is the time of the silent part obtained by the second decoding means in between
Signal output means for inserting a silent signal.
Voice decoding device.