JPS5994935A - Control circuit of voice interpolation - Google Patents

Abstract

PURPOSE:To suppress a noise due to burst error without deteriorating articulation by gating a burst error detecting signal in response to its period to form an interpolation control signal. CONSTITUTION:A varible length pulse generator 10 outputs a gate signal 55 in synchronization with a frame synchronizing signal 58. An AND circuit 11 ANDs the burst error detecting signal 54 and a signal 55 and outputs the result as an interpolation control signal 56. A decoder 6 possible for interpolation in the unit of frames performs normal decoding by using a coding voice signal 57 when the signal 56 does not go to 1, and the pitch synchronism interpolation is attained by using the information of a frame before that going to 1. Thus, the pitch synchronism interpolation is attained in the unit of frames for the error of an important bit and the comparatively light bit error is neglected. As a result, the deterioration in articulation due to excessive interpolation is prevented and the noise due to burst error is suppressed.

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to digital voice communication using a transmission path where burst errors occur, and more particularly to a decoder control device capable of interpolating decoded voice on a frame-by-frame basis.

Conventionally, decoded voice interpolation as a technique for preventing voice quality deterioration when burst errors occur has been applied to voice codes and decoders that quantize voice sample values as they are, such as PCM. In this method, there is a one-to-one correspondence between the section where burst errors occur and the decoded speech section where the decoded speech deteriorates due to burst errors, so the burst error detection signal is used as an interpolation control signal, and the decoder receives the interpolation control command. It would be sufficient to perform interpolation only on the extracted sections. However, Hurray.

A highly efficient speech coder and decoder that performs coding and decoding by dividing the code into multiple frames (e.g. adaptive bit allocation prediction (APC-AB);
Adaptive Predictive Codin
g with Adaptive Bit Al 1o
In cat 1on)), there is not necessarily a one-to-one correspondence between the section where the burst error occurs and the decoded audio section where the burst error degrades, and the audio of the entire frame may be affected, so decoding is performed frame by frame. It is necessary to perform audio interpolation. If the burst error detection signal is used as it is as an interpolation control signal for such a decoder, even if the bit error rate is the same, the frame error rate (
When the number of frames containing errors/all frame changes are different, when the period of burst errors is sufficiently long compared to the frame length, the frame error rate is small and good interpolation can be performed. When the length is shortened, the frame error rate increases and noise caused by burst errors can be suppressed, but the clarity of decoded speech deteriorates due to excessive interpolation.

The present invention has been made in order to eliminate the above-mentioned drawbacks of the conventional technology, and an object of the present invention is to adaptively change the time length of the gate signal depending on the period of burst errors.
The frame interpolation rate (number of interpolated frames/total number of frames) is kept below a certain value (for example, Vlo) determined by the bit error rate of the transmission path, regardless of the cycle of burst errors, and burst errors with a cycle that is sufficiently long compared to the frame length are However, for burst errors with a short cycle compared to the frame length, interpolation is performed only when an error occurs in a bit that is important for waveform distortion. An object of the present invention is to provide a new device for controlling a decoder that enables frame-by-frame interpolation so as to maintain high clarity and suppress noise caused by burst errors without complicating the decoder.

In order to achieve the above object, an audio interpolation control circuit according to the present invention includes a first device that outputs a burst error detection signal indicating that a burst error has occurred on a transmission path, and a first device that outputs a burst error detection signal indicating that a burst error has occurred on a transmission path, and a first device that decodes audio in each frame. By outputting a gate signal that becomes '1' in an interval of n bits, and adaptively changing the value of n depending on the period of burst errors, the frame interpolation rate can be adjusted to the bit error rate of the transmission path. and a third device that uses an AND of the burst error detection signal and the gate signal as an interpolation control signal for a decoder that performs frame-by-frame interpolation. configured.

Next, a preferred embodiment of the present invention will be specifically explained with reference to the drawings.

FIG. 1 is a block diagram showing an embodiment of the present invention. In FIG. 1, reference number 1 is an A-D converter, 2 is an adaptive bit allocation prediction (PC-λB) encoder, and 3 is a modulator. ,
4 is a transmission path, 5 is a demodulator, 6 is an APC-AB decoder capable of frame-by-frame interpolation, 7 is a D-large converter, 8 is a detector, 9 is a threshold circuit, 10 is an extended variable length pulse generator , 11 indicate the 0 circuit, respectively.

The detector 8 and the threshold circuit 9 constitute a first device for detecting burst errors in the transmission path, and the variable length pulse generator 10 synchronizes with frames of audio divided into fixed time lengths and detects the burst errors in the transmission path. A second device is configured to output a gate signal whose time length υ changes depending on the period of the burst error detected by the first device,
A third device is configured to input the burst error detection result by the device and the gate signal output by the second device and output an interpolation control signal.

As shown in FIG. 1, an input audio signal 51 is converted into a digital signal by an A-D converter 1, divided into 16 msec frames by an APC-AB encoder 2, and then converted to a 16 Kbps signal.
The bits are output as an encoded audio signal 52 in order of importance in decoding the audio. The signal 52 is
The signal is modulated by a modulator 3, passes through a wireless transmission path 4 where aging occurs, and becomes an input to a demodulator 5 and a detector 8. The demodulator 5 outputs an encoded audio signal 57 containing burst errors. On the other hand, the signal input to the wave detector 8 is converted into a received power signal 53 by the wave detector 8 in order to detect burst errors, and is further input to the threshold circuit 9 so that the signal 53 exceeds the threshold value. Sometimes it is 10", and when it is less than 1", it is output as the burst error detection signal 54. Variable length pulse generator 1
As shown in Figure @2, 0 is synchronized with the frame synchronization signal 58, and for a constant bit error rate, it is always @1" when processing is sufficiently slow compared to the frame period, and becomes shorter as fading becomes faster. The AND circuit 11 outputs a pulse as the gate signal 55.
5 is ANDed and output as an interpolation control signal 56.

The APC-AB decoder 6 receives the signal 5 even once within the frame.
If the signal 56 does not become ``1'', normal decoding is performed using the signal 57, and in frames where the signal 56 becomes ``1'' at least once, the information of the previous 7 frame is used. The digital audio decoded or interpolated by the decoder 6 is converted by the D-A converter 7 and output as decoded audio 59.

Fig. 2 shows an example of the signal waveform of each part. During 7 aging which is slower than the frame period (Fig. 2 @), the gate signal 55 becomes '1' in all sections within the enlarged frame, and the burst error detection signal 54 is interpolated as is. Since the control signal 56 is generated, noise due to burst errors is not output to the decoded voice.When aging is faster than the frame period (Fig. 2 (A)), the gate signal 55 is generated by controlling the more important n bits in the frame. It becomes @1'' at the position and 0'' at other positions. Therefore, pitch-synchronized interpolation is performed on a frame-by-frame basis for important bit errors, and ignored for relatively light bit errors.As a result, Although noise due to burst errors is output to the decoded speech, it is rarely small, and deterioration of clarity due to excessive interpolation can be prevented.

Although the configuration and operation of the present invention have been described above with reference to one preferred embodiment thereof, this is merely an example, and the present invention may include other aspects without departing from its scope.
It can be implemented with various modifications and changes.

As explained above, the present invention gates the burst error detection signal in accordance with the period of the burst error to generate an interpolation control signal. This has the effect of suppressing noise due to burst errors 9 without significantly deteriorating clarity.

[Brief explanation of the drawing]

@Figure 1 is a block configuration diagram showing one embodiment of the present invention, Figure 2 is a block diagram showing an embodiment of the present invention.
The figure is a diagram showing an example of signal waveforms at various parts in FIG. 1. 1.-@A-D modulator, 2.. Adaptive bit allocation prediction (APC-AB) encoder, 3.. Modulator, 40 threat/transmission path, 5.. Demodulator, 6.. - APC-AB decoder capable of frame-by-frame interpolation, 7 - DA converter, 8
- Detector, 9... Threshold circuit, 10... Variable length pulse generator, 11... 0 circuit, 51...-Human voice signal, 52... Encoded voice signal, 53... ...Received power signal, 54...Burst error detection signal, 55...Gate signal, 56...Interpolation control signal, 57...Encoded audio signal including error signal, 58...Frame synchronization signal , 5
9. Decoded audio patent applicant Yuhato Kumagai, agent for NEC Corporation, patent attorney

Claims (1)

[Claims] An audio encoder that divides audio into frames of a fixed time length and encodes it; and a decoder that, when an interpolation control signal is input, interpolates the decoded audio of the frame that includes the input point with the decoded audio of the previous frame. In a voice communication system using a device, a first means for detecting a burst error in a transmission path;
a second means for outputting a gate signal whose time length changes depending on the period of the burst error detected by the first means in synchronization with the frame; tR3 means for inputting the gate signal from the second means and outputting the interpolation control signal.