JPS58111991A - Voice reproducer - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

DETAILED DESCRIPTION OF THE INVENTION Technical Field of the Invention The present invention relates to an audio reproduction device that reproduces an audio signal sound from audio signal data encoded by a method such as delta modulation.

Technical background of the invention and its problems Recently, audio recording and playback systems have been put into practical use that encode audio signals and store them in memory for use in various guidance services, language training, etc. This system is particularly effective. In the case of consumer products, there is a strong desire to lower the data rate and store data even if some naturalness is sacrificed due to cost considerations, and PCM, DM (
No matter which encoding method is used, such as delta variable (1111) or LPC (linear predictive coding), if the bit rate is lowered, consonants in particular will become unclear, and the deterioration of intelligibility will be minimized. 4
Although the delta modulation method, which allows encoding at a lower pit rate than IiK, is promising for telecommunications, it has the following problems.

The delta modulation method basically samples the audio signal waveform at a cycle set short enough to keep the difference between adjacent Funzol values within a certain range, and only the magnitude relationship between adjacent Funzol values is determined. It is expressed as data and quantized.

Figure 1 is a waveform diagram showing the principle of the tube.The input signal waveform in (a) is first converted into a staircase waveform as shown in (b). This is the output waveform of the local output signal (-), and the output waveform of the decoder is also the same.

And the waveform of (2) shows (e) K11′″@Qll
It is transmitted or stored in memory as a data string.

The biggest advantage of the NiO delta modulation method is that it is relatively simple.
- Although it is possible to encode trades using hardware with relatively low power, on the other hand, its disadvantage is that its flexibility is low. One phenomenon with low fidelity is Datsuni,
- This is called noise, and in Figure 1 (the small line of the input signal waveform of Xun) or the no-signal part 11, alternating K @1"@lQ# occurs in the differential data, and the local D of (B) /A controller/4-Yu's output waveform oscillates at the Nanzoring period, ←
This is noise caused by @1m and @QTa appearing alternately in the encoded data of ). The other type is called O/G-slope load, which is caused by the local output waveform of the converter being unable to follow the changes in the steeply sloped part of the input signal waveform shown in Figure 1 (a). This is the distortion that occurs.

As a method to compensate for these shortcomings, if the difference data of the input signal waveform is alternately @11'' and @0'', the quantization step width in the local V-converter output waveform can be adjusted to reduce grani and -ra noise. Gradually reduce the difference, and if the difference numerator - e is, for example, 3 to 4 or more consecutive "1" ma9 becomes "0", conversely, gradually increase the quantization step and deconvergence to exceed An adaptive lid delta modulation method that reduces the slow fO-load has already been put into practical use, and has shown improvements in the following.However, another large distortion occurs in the delta modulation method, including the adaptive type. This is welfare! It is like a beat between an integer fraction of the frequency and the input signal frequency, and becomes a large noise when the sampling frequency is relatively low. For example, when the sampling frequency is 16 kHz,
Figure 2 (1 kHz of steep slope (large level) as shown in -)
If the input value of z) can arrive, the output waveform of the local V mufunpa voice will be k as shown in (b), and therefore the encoded data will be 1
There are eight consecutive -1''' and @O' as shown in (・).This means that the input signal frequency is 1 of 1 kllx.
6/1. 16 kHz K, that one is 1s
It must be expressed in bits, in which case, for example, the local D/
The output waveform of the A humperter is the ssth (b), and the encoded data is as shown in (2). K @1" and @O"I are 8-7-
7-8 bits, G and so on, each wave is distorted greatly, and the input signal frequency is 1 kHz.
At 1.082 kHz of 31, the output waveform and encoded data of the local i pumper are shown in Figure 4 (&), (
As shown in b), when playing these distortions, which are greatly distorted every two waves, especially when the replay noise is comparatively large, it appears as a background noise that is extremely harsh on the ears.

Conventionally, as a countermeasure to this phenomenon, the high frequency of the demodulated waveform is cut using an output filter whose cutoff frequency on the high frequency side is relatively low. However, with such an output filter the high frequencies are suppressed.

When the background noise is played back, the background noise becomes more noticeable, but on the other hand, information on consonant parts with many high-frequency components is lost, which greatly impairs the clarity of the reproduced speech. The purpose of this is to develop an audio playback device that can reduce background noise without compromising clarity when reproducing the original audio from data that has been encoded using the delta cost adjustment method. It is about providing.

SUMMARY OF THE INVENTION This invention demodulates the audio signal by receiving encoded audio signal data together with a hood for discriminating vowel parts and consonant parts and a consonant part O duration code, and at the same time demodulating the audio signal. cascade the filter and amplifier on the side,
Discriminating the consonant period of the audio signal data from the discrimination code and the duration code, and emphasizing the consonant part of the audio signal output from the tube and decoder during the period, at least one of the passband width of the Funi filter and the gain of the amplifier. The feature is that it is controlled in the direction of

Effects of the Invention According to this invention, high-frequency components that normally contain a lot of background noise are suppressed by a filter, and consonant parts are emphasized by widening the passband width of the filter or increasing the gain of the amplifier. Therefore, the intelligibility can be improved.In this case, the consonant part is originally at a low level, the ninth,
The background noise mentioned above is relatively inconspicuous, so there is no problem in emphasizing this part. That is, according to the present invention, when reproducing round audio signal data encoded at a low bit rate, Background noise jPt--can be effectively reduced without compromising clarity.

Embodiment of the invention Schematic configuration 11 of an embodiment of the invention of gsIiI#ijin
11, the memory 51 stores coded voice signal data (for example, by delta modulation), a discrimination code for distinguishing between a vowel part and a consonant part, and a duration indicating the duration of the consonant part. It is stored together with the moxibustion hood, and its contents are sequentially read out under the control of the controller sxi. ? :,
The output of memory J1 is led to decoder IRK, where the audio signal is demodulated. This audio signal is applied to the speaker 51 via a filter 55 and an amplifier 56t in sequence.
Played as audio. On the other hand, the output of the memory 51 is also given to a consonant period discrimination circuit JILK, which discriminates the consonant period based on the output KtttLJt of the memory 5, the discrimination hood between the vowel part and the consonant part, and the duration hood of the consonant part. Ru. According to the output jjK of this consonant period discrimination circuit 54,
The passband width of filter 56 and the gain of amplifier 51 are controlled.

The filter 56 is, for example, a low-pass filter whose passband width can be switched in two stages, and an example of its structural configuration is shown in FIG. electronic switch 61.6
4 and 10 by turning on these switches 63 and 64.
A 65° resistor connected in parallel to the circuit gxo resistor section can be added, and when switch ill and 64 are turned on,
High frequency side force, to-off frequency from fl to /s (/@<f*
).

In this ζ, /*, /s are selected to be, for example, /l-1.2kHz, fwm2.4kHz@degree.

On the other hand, the amplifier @11 is a variable gain amplifier whose gain can be switched in two stages.For example, as shown in FIG.
Input resistance 1 of the inverting amplifier consisting of FJ and resistors vx and yx
2 in parallel with the electronic switch F4′f: via the resistor raY
t are connected in parallel, and when the switch 14 is turned on, the resistance value of the input resistor decreases and the gain increases.

Is the output of the consonant interval discrimination circuit 114Fi and the memory 51, for example, 8 bits)? It is in parallel data format, and the discrimination code between the vowel part and the consonant part and the duration length food of the consonant part are outputted before the voice data, and each of these codes is output as the upper 2 bits, g, and lower 6 bits, respectively. If it is expressed in bits, it can be realized with a configuration as shown in FIG. 8, for example.

In FIG. 8, the compiler 81 compares the 2-bit vowel part and consonant part from the memory j1 with the hood from the code generator 11 and the code IPan from the code generator 11. The negative logic signal a4t- of the width of the evening period is output. However, the punno arm regulator a1 is controlled by the signal 11 from the controller j1 in FIG. No, counter 8C is memory 5 at the down edge of this no 4rus a4.
Loads the 6-bit duration code JIFi of the consonant part from 1 and enters down count mode ・Meanwhile, Noculus 8
At the up edge of 4, 8-R7 and Zoflofa1 are set, and its Q output becomes the output of the consonant period discrimination circuit s4. This Q output sets the counter a6t-enabled and starts a down-counting session. Counter 8e (by this down count) when its contents become zero, Ichiga Hajime/
4 pulses are output, and 7 reset fax'fI: is set in the reset state, and at the same time, 1 itself also becomes in the tisezol state. In this way, a high level output 15 is obtained at the Q output of f 7 a y 7' II J during the consonant period, and this turns on the switch 1 and #4 in FIG. At the same time, switch 14 in FIG. 7 is turned on to increase the gain of amplifier j4.

As above, %ζOm1! According to the embodiment, the filter 16 is normally set to have a narrow band shape 11 with a relatively low cutoff frequency on the high side, thereby suppressing the generation of background noise. On the other hand, during the period when the data of the consonant part is output from the memory 11, the filter j# moves to a higher cutoff frequency on the high frequency side and enters the wide range state, and at the same time, the gain of the amplifier i1 changes. The intelligibility is effectively improved by increasing the consonant part and emphasizing the consonant part. ・If low pitch and pitch are required when encoding the vocal pronunciation code, the delta modulation method as mentioned above is used. is often used, but the ensuing method of this method usually has a frequency characteristic tube with extremely limited high range.For example, as shown in Fig.
l@81op* Delta ybduration
a) Audio recording and playback system (16kb/s'
) The frequency characteristics of iD modulation and demodulation are shown as a meter with an input level, so it has a strong electric current effect that depends on the frequency. That is, for example, if we focus on 2 kHz, the resulting output level will be -20d as we increase the input level.
B It does not exceed about (0,14Vrml'),
Therefore, the present invention is highly effective when reproducing audio signal data encoded by such a method.

Note that in the above embodiments, both the passband width of the filter and the gain of the amplifier were controlled during the consonant period, but the effect of improving intelligibility can be obtained even if only one of them is controlled. In addition, the encoding method is the delta modulation method (
It may also be based on other methods such as PCM, LPG, etc. (including adaptive type). This invention is effective in case of 'lK low vi, trade. However, in the case of the PCM method, the passband width of the filter is set to 4, the frequency is set to 1/2, and only the gain of the amplifier is controlled.
I love it.

[Brief explanation of drawings]

Figure 2 is a waveform diagram to explain the principle of the delta modulation method, Figure 2 is a waveform diagram when the input signal frequency is an integer fraction of the sampling frequency, Figures 3 and 4 are waveform diagrams when the input signal frequency is a sampling frequency. A waveform diagram when the frequency is not an integer fraction, FIG. 6 is a diagram showing the schematic configuration of an audio reproduction device O according to an embodiment of the present invention, and FIG. 6 is a diagram of a variable passband width filter in the embodiment. Diagram 7 showing a specific configuration example
Figure 8 is a diagram showing a specific configuration example of a variable gain amplifier, Figure 8 is a diagram showing a specific configuration example of a consonant interval discrimination circuit, and Figure 9 is a diagram showing an example of modulation/demodulation envelope frequency characteristics of the CVgD method. be. 11J... memory, 52... controller, 11
- Decoder, 54...Consonant section discrimination circuit, 1#--Variable pass band width filter, 61... Variable gain amplifier, S
a...Speaker. Applicant's agent Patent attorney Suzue Takehiko Shin 1- Figure 2 (c) 8 8 8 8 Figure 3 Figure 4 (b) 8878878 1511 Figure 6 811 9th WR rilejkl,, (H2)

Claims (1)

[Claims] A decoder receives encoded large audio signal data together with a vowel part and consonant part discrimination hood and a consonant part duration length hood and demodulates the audio signal, and an output side KI1 of this decoder.
1, a filter and an amplifier connected in series to the filter, and means for reproducing sound through the filter and amplifier;
The consonant period of the audio signal data is discriminated from the discrimination boolean and the duration code, and during the period, at least one of the passband S of the filter O and the gain of the amplifier is set to the consonant period of the audio signal output from the debugger. 1. A sound reproduction device comprising a consonant period discrimination circuit for controlling in a direction of emphasizing a consonant period.