JPS6037849A - Voice signal transmission system - Google Patents

Abstract

PURPOSE:To improve the line utilizing efficiency by setting initially a forecast coefficient and a step size for a coder and a decoder of a channel to which no channel is allocated. CONSTITUTION:A reception side multiplexer 40 is changed over in synchronizing with a transmission side multiplexer 36, a signal subject to time division multiplex at a line 38 is shared to channels CH1, CH2..., and the decoder 42 converts the signal into an analog signal. When a no-talking detector 34 detects the absence of talking, the coder 30 and the decoder 42 reset the coefficient of the forecast device to ''0'' and also minimize the step size of a quantizer.

Description

DETAILED DESCRIPTION OF THE INVENTION Technical Field of the Invention The present invention relates to an audio signal transmission system that combines DSI and ADPCM.

Prior Art and Problems Telephone voice signals usually have interruptions due to various reasons such as breaks between words, not being able to find the next word, waiting for a reply from the other party, etc. However, from the point of view of the communication line, it is a waste of time because it is idle during the interruption period. Therefore, if the interruption period is used for transmitting audio signals of other channels, a high link ratio can be achieved. The transmission method is DSI (Digi
tal 5peach 1nLerporation)
It is. FIG. 1 is a diagram explaining this, where ta+ indicates the audio signal of channel 1, (bl indicates the audio signal of channel 2, and both have interruptions or idle periods T1 to 1゛3.(C1
shows the signal waveform when the audio signal of channel 2 is sent during the idle period T1 of channel 1, and Fl and 'F2 are status flags. On the receiving side, the audio signal is separated by this status flag and returned to the original channel 1.2 signal. By using the interruption period to send audio signals of other channels in this way, lines can be used effectively, and this is appropriate when using expensive lines such as submarine cables and satellite communications.

Additionally, LRE (Low Rate) is used to transmit audio signals.
There is an encoding method called IEncoding. Encoding of energization is performed at 64, etc., but in LRE, it is performed at 32, 16, 2.4, etc. ADPCM, which is effective for band compression, is one of them. AD
Live Differential) PCM is a DPCM that changes the quantization step and/or coefficients of the predictor when the difference becomes significant. FIG. 2 shows its encoder, and FIG. 3 shows its decoder. In FIG. 2, S tn+ is an input audio signal, which is subtracted from the predicted signal P (nl) by a subtracter 16, and the difference E tn+ and the output V (
The product 1Esfn) with n) is input to the non-uniform quantizer 10 and sent out as a quantized signal 1 (nl). The predicted signal P (n) is generated by the inverse quantizer 14, etc. The quantized output I (n) is input to the inverse quantizer 14 and is converted into an analog signal Es (nl) corresponding to the input △ of the quantizer 10, and is the product of this and the output △fn of the adaptive logic circuit 12. The sum of E (n) and the prediction signal P (nl (this corresponds to S (nl)) is stored in the memory element 20a. There are multiple memory elements 20a to 20d, and they are sequentially shifted using the sampling clock. and coefficients a1 to a are applied to the output of each memory element.
The adder 22 calculates the sum of the products multiplied by 4. The output of this adder 22 is a prediction signal P (n), which in this example is obtained by multiplying each audio signal from the current time up to four sampling times ago by coefficients a1 to a4. ADP
In CM, if the difference between the predicted signal Pfnl and the input signal S(nl is large (the prediction is incorrect), the coefficients a1 to a4 are modified to reduce the error, and the quantization step is changed.

Decoding is the opposite of encoding, and as shown in FIG. 3, the quantized signal I (nl) from the transmitting side is
n), and is created by the adaptive logic circuit 26 and is converted to Δ(・
) is input to the adder 28, and the predicted signal P(
This becomes the input audio signal S tn + phase △ corresponding output audio signal S (nl. Prediction signal P (n
l is created in the same way as on the sender side. Note that g is a sofa amplifier. What is important in decoding on the receiving side is the coefficient a1 of the predictor.
~a4, quantization step, etc. are the same as on the transmitting side.

When DSI is applied to ADPCM, the problem is how to handle channels whose transmission is interrupted, because DSI switches the transmission channel when there is no signal.

ADPCM is based on the correlation of audio signals at each point in time, but if the signal is interrupted, the correlation is lost and accurate prediction signal generation cannot be expected.

OBJECTS OF THE INVENTION The present invention seeks to address the problems associated with DSI
The aim is to operate AD and PCM effectively and further improve the J rate of line usage.

Composition of the Invention The present invention provides an audio signal transmission system in which audio signals of ADP and CM are transmitted using DSI, and in a transmitting side ADPCM encoder and a receiving side ADPCM decoder of a channel to which a communication path is not assigned, predictor coefficients are The present invention is characterized in that the step size and step size are initially set to predetermined values, which will be explained next with reference to embodiments.

Embodiment of the Invention FIG. 4 shows an ADPCM audio signal transmission system to which DSI is applied, and 30 is the ADPCM encoder shown in FIG.
,... are provided respectively. 32 is D
The SI includes a no-call detector 34, a multiplexer 36, and the like. When the detector 34 detects a non-call, the multiplexer 36 switches the busy channel to another busy channel so that the other busy channel is connected to the communication line 38 and its signal is transmitted to the receiving side. . There is a multiplexer 40 and an ADPCM decoder 42 on the receiving side, and the multiplexer 40 switches in synchronization with the decoder 42 on the transmitting side, and the line 38 sends the time-division multiplexed signals to each channel CHI, CH2°, etc. ...distributed separately,
Decoder 42 converts it to an analog audio signal.

If the channel is switched without a call, the transmitting and receiving sides of the channel that were previously connected by line will be disconnected, and the predicted signals of the two will deviate greatly. Therefore, in the present invention, when switching channels, the switching signal is received from the DSI 32, and the A
The DPCM encoder 30 and decoder 42 use the predictor coefficient a
1 to a4 are reset to 0 and the quantizer step size is minimized. FIG. 5 is a diagram explaining this. When the detector 34 detects silence (no call), the detection signal is used for channel switching as described above, and is also sent to the previously connected ADPCM coder and decoder. Enter to change predictor coefficients and step size. 6th
This is a diagram illustrating step size change on the transmitting side, in which the adaptive logic circuit 12 sends a signal to the m-digitizer 10 and the inverse quantizer 14 to change the quantization step size using a differential quantization signal I (b). is generally composed of 4 bits, but the difference can be large or small, so the size (step size) represented by each bit is changed to deal with large or small differences. In this way, as is clear from FIG. 2 etc., the predicted signal P (
nl becomes 0, and the input audio signal S (nl remains as it is\
It is quantized (not a difference) and sent out, and decoded on the receiving side. At the next sampling point, △ alEfnl becomes the predicted signal P (nl, and the difference from this is quantized and sent out. The same goes for the rest, and in the ADPCM of FIGS. 2 and 3, the normal state is returned after 4 samplings. During this time, since the predicted signals on the transmitting and receiving sides have the same value starting from 0, decoding is performed correctly.

When restarting transmission after an interruption, the audio signal often gradually increases from a small amplitude. Resetting the step size to a minimum upon interruption of transmission can better accommodate such characteristics.

When resetting ADPCM at the time of interruption, instead of setting all coefficients a1 to a4 to 0, al=1. It may be a2 to a4-〇. In this case, the predicted signal when transmission is restarted is the signal sent immediately before the interruption, and the difference between it and that signal is sent and decoded when transmission is restarted. al=l.

a2 to a4=0 are DPCM, and this example is ADPC
While the M encoder and decoder are suspended, they will be locked to the DPCM algorithm ↓. The original method has better tracking characteristics than the method in which the coefficients a1 to a4 are all O.

If the initial stage is set to all θ, etc., and in the example shown in FIG. 2, a1 to a4 are set to 0000 or 1000, and the state of the center concave + iij is not maintained, there are the following advantages. That is, DSi
As shown in FIG. 7, the ADPCM coder 30, DSi (
For details, see the multiplexer) 32, line 38, DSi
(Same) 32, the configuration of the ADPCM decoder 42 (
(This is the same as in Figure 4, etc.) Instead, D as shown in Figure 8.
Si32, ADPCM coder 30, line 38, ADPC
It is also possible to configure the M decoder 42 and DSijO, which can reduce the number of cogs and decoders. Therefore, in the example of FIG. 7, the 24 channels from chi to ch24 must be encoded and decoded by the cog 30 and the decoder 42, respectively, and 24 cogs and decoders are required, but the method shown in FIG. Preferably, the 24 channels from ch1 to ch24 are combined into 12 channels by the DSi and encoded and decoded, so the number of cogs 30 and decoders 42 may be 12 each. However, in this method, the relationship between the person, the output side channels chi to ch 24, the cog 30, and the decoder 42 is not fixed, and the DSI cuts off the channel that has become non-calling and connects the communication channel in its place, so each cog , a decoder is arbitrarily assigned to each channel (the assignment method is determined by the DSI), and the memory of the state before the interruption becomes meaningless. This point oooo or 1000
Initialization to is valid.

As described in detail, in the present invention, since the ADPCM audio signal is sent by DSI, more effective use of the line can be achieved.
When switching channels in SI, the predictor coefficients and step size of the ADPCM encoder and decoder of the channel to be switched are preset, so that when the channel is restored, the predicted signal on both the transmitting and receiving sides rises correctly from the same magnitude, has good followability, and is accurate. Decryption can be resumed.

[Brief explanation of drawings]

Figure 1 is an explanatory diagram of DSI, Figures 2 and 3 are ADP
Block diagrams of the CM encoder and decoder, FIGS. 4 to 6 are block diagrams showing embodiments of the present invention, and FIGS. 7 and 8 are block diagrams of the CM encoder and decoder.
The figure is an explanatory diagram of a modified example. In the drawing, al to a4 are predictor coefficients, and MUX and DMUX are channel switching devices. Applicant Fujitsu Limited Representative Patent Attorney Minoru Aoyagi

Claims (1)

[Claims] In an audio signal transmission system that sends an ADPCM audio signal using DSI, the transmitting ADPCM encoder and the receiving ADPCM decoder for channels to which no communication path is assigned use predetermined values for the predictor coefficients and stave size. An audio signal transmission method characterized by initial settings.