JPS6336169B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION (1) Technical Field of the Invention The present invention relates to a voice compression and demodulation method, and particularly to a method for transmitting a telephone voice signal by compressing its frequency bandwidth to a small value.

(2) Background of the Invention In order to efficiently use a dedicated telephone line or the like, a method of dividing and multiplexing the frequency of a transmission signal to enable calls over multiple channels is well known. In this frequency division multiplexing method, it is necessary to compress the frequency band of the audio signal on the transmitting side and demodulate it to the original frequency band on the receiving side. An excellent method is to compress and transmit audio by extracting certain frequency components from the audio signal and inserting other frequency components into the extracted portions. In this case, the apparatus for compressing and demodulating audio signals is required to have a high performance filter for removing unnecessary frequency components and to be inexpensive.

(3) Conventional technology and problems Conventionally, an analog comb-type filter has been used as a filter to filter out unnecessary frequency components. Also, in order to put other frequency components in place of the extracted frequency components, or conversely to separate them,
The amplitude was modulated using a sine wave with a frequency of F/2, but
In this conventional system, the performance of the comb filter is insufficient, and since a sine wave generation circuit and an amplitude modulation circuit are required, the number of components is large, and the cost of the device is high.

(4) Purpose of the Invention In view of the above-mentioned problems in the prior art, the purpose of the present invention is to realize a filter with a high-performance and simple digital filter, and to thin out the sample value of the digital signal instead of amplitude modulation using a sine wave. Based on the concept of inserting a sample value of zero value between each adjacent sample time,
An object of the present invention is to reduce the number of parts required for a device in an audio compression and demodulation system, thereby reducing the cost of the device.

(5) Structure of the invention The gist of the present invention to achieve the above object is as follows:
When the highest frequency of the input audio signal is F, on the transmitting side, the input audio signal is set to the sampling frequency.
A first analog-to-digital converter that performs AD conversion at 2F, and waves the output signal of the first analog-to-digital converter so that the pass band and stop band are located at symmetrical positions with respect to frequency F/2. A first digital filter, by thinning out the output signal of the first digital filter at every sample time, a signal component from frequency zero to F/2 and a signal component from frequency F/2 to F are separated from each other in an empty frequency range. a sample thinning circuit that returns the sample to
The receiving side includes a second analog-to-digital converter that AD converts the analog signal transmitted from the transmitting side at a sampling frequency F, and converts the output signal of the second analog-to-digital converter. During each adjacent sample time,
A zero sample insertion circuit that inserts sample values of zero value one by one and outputs them at a sampling frequency of 2F;
having the same characteristics as the first digital filter,
a second waveform that waves the output signal of the zero sample insertion circuit;
digital filter, and the output signal of the second digital filter at a sampling frequency of 2F.
The audio compression and demodulation method is characterized in that it includes a second digital-to-analog converter that performs DA conversion and outputs the result to a receiver.

According to one aspect of the present invention, the first and second digital filters each include a delay element included in a digital low-pass filter or a digital high-pass filter with a sampling frequency of 2F and a cutoff frequency of F/2. N pieces (N is an odd number)
The configuration is such that the delay elements are replaced by cascade-connected delay elements.

(6) Examples of the invention Embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram showing a transmitting section that compresses an input audio signal according to an embodiment of the present invention, and FIG. 2 is a block diagram showing a receiving section that demodulates a received signal.

In Figure 1, 1 is a transmitter that outputs an audio signal with a maximum frequency of 4KHz, and 2 is a first analog-to-digital converter (hereinafter referred to as AD) that converts the input audio signal from the transmitter 1 into an AD signal at a sampling frequency of 8KHz.
3 is a first digital filter, 4 is a sample thinning circuit, and 5 is an output signal of the sample thinning circuit at a sampling frequency of 4KHz.
A first digital-to-analog converter (hereinafter referred to as DA converter) performs DA conversion, and 6 is a transmission line.

The analog audio signal input from the transmitter 1 to the first AD converter 2 has, for example, the analog spectrum shown in FIG.
AD conversion is performed at a sampling frequency of 8KHz, which is twice the highest frequency above, and the same result is shown in Figure 5a.
The result is a digital signal with a digital spectrum shown in . However, since it is a digital signal, it is 4K.
There is a spectrum above Hz, but the same one as the digital spectrum between 0 and 4KHz is repeated, and to simplify the drawings, the spectrum above 4KHz is omitted in Figures 3 to 5. . The output of the AD converter 2 is waved by a digital filter 3. As shown in FIG. 3, the digital filter 3 has a wave characteristic in which a pass band and a stop band are located at symmetrical positions with respect to a frequency of 2 KHz, which is 1/2 of the maximum frequency of 4 KHz. So, 0~1KHz, 1.5~2KHz, and 2.5~3KHz are the passbands, and regarding 2KHz
1-1.5K located symmetrically with the 2.5-3KHz passband
Hz, and 2 to 2.5 KHz, which is located symmetrically to the 1.5 to 2 KHz pass band with respect to 2 KHz, is the stop band. What is important for audio signals is the vowel frequency band of 0 to 1 kHz, which determines the naturalness of the sound, and even if the frequency band is passed through a filter with the characteristics shown in FIG. 3, the quality of the audio will not deteriorate significantly. In this way, a signal having the spectrum , and as shown in FIG. 5b is obtained at the output of the digital filter 3.

The output signal of the digital filter 3 is a digital signal sampled by the AD converter 1 at a sampling frequency of 8KHz.
As shown in Figure a, sample values are provided at every 125 μS time interval. This is processed by the thinning circuit 4,
The signal is thinned out at every sampling time and converted into a digital signal having a time interval of 250 μS as shown in FIG. 4b. This makes the sampling frequency
It becomes 4KHz which is half of 8KHz. When a digital signal is thinned out every sample time in this way, an aliasing phenomenon appears due to the sampling theorem, as is well known, and the frequency spectrum becomes 2K at the center frequency of the input audio signal.
The signal is folded back symmetrically with respect to Hz, resulting in a signal having a spectrum as shown in FIG. 5c. In Figure 5c, the signal in the 2.5-3KHz band of Figure 5b is
It can be seen that the signal is folded back to a band of 1 to 1.5 KHz, which is a symmetrical position with respect to 2 KHz. In the spectrum shown in FIG. 5c, the frequency relationship is reversed from that in the case where the output of the digital filter 3 is amplitude-modulated with a 2KHz sine wave.

The output signal of the sample thinning circuit 4 is converted into an analog signal by a DA converter 5 at a sampling frequency of 4 KHz and sent to a transmission line 6.

In FIG. 2, 7 is a receiving transmission line, 8 is a second AD converter that performs AD conversion at a sampling frequency of 4KHz, 9 is a zero sample insertion circuit, and 10 is a second digital filter having the same characteristics as the first digital filter 3. Digital filter, 11 is sampling frequency 8K
a second DA converter that converts DA at Hz, and 12
is the receiver.

Second AD converter 8 via transmission lines 6 and 7
The analog signal input to is the sampling frequency
It is converted to a digital signal at 4KHz. The output signal of the AD converter 8 has the spectrum shown in FIG. 5c, similar to the output signal of the thinning circuit 4.
Since it is sampled at 4KHz, the sample time interval is 250μS. A zero sample insertion circuit 9 is inserted between each adjacent sample time of this signal.
As a result, sample values with a value of zero are inserted one by one, and the sample time interval is set to 125 μS. In this way, the sampling frequency of the output signal of the zero sample insertion circuit 9 becomes 8KHz, and its spectrum becomes
The result will be as shown in Figure d. The spectrum in Figure 5 d is the spectrum in Figure 5 c at a frequency of 2KHz.
The symmetrically folded spectrum is synthesized with the original spectrum and has a frequency band of 4KHz, which can be explained as the folding phenomenon of the sampling theorem. The output signal of the zero sample insertion circuit 9 is waved by a second digital filter 10 having the amplitude characteristics shown in FIG.
A signal having the spectrum shown in Figure b is demodulated. This is converted into an analog signal at a sampling frequency of 8 KHz by the DA converter 11 and output to the receiver 12.

In this way, according to this embodiment, the folding of the spectrum and the reverse operation can be realized with an extremely simple circuit.

Note that, like digital filters 3 and 10, on the frequency axis of frequencies 0 to F, the frequency F/2
A method of realizing a digital filter with a sampling frequency of 2F (this will be called an antisymmetric filter) in which the passband and stopband exist at frequency positions that are symmetrical to each other around the point, with fewer calculations. Conventionally, there has been a comb-shaped digital filter having a sinusoidal amplitude characteristic, but this has the drawback that the boundary between the pass band and the stop band is not clear. Therefore, the filter output from 0 to
When the F/2 component and the F/2 to F components are put into the empty frequency range by modulation, adjacent passband parts overlap each other (this phenomenon is called interband crosstalk). ), the audio quality during demodulation is likely to deteriorate.

In order to eliminate the above-mentioned drawbacks of the conventional comb-shaped digital filter, the present invention uses a digital low-pass filter with a sampling frequency of 2F and a cutoff frequency of F/2, or a digital high-pass filter (this is called a half-band filter). ) is first designed and the circuit configuration is determined, and then each delay element included in this circuit is replaced with N (N is an odd number) cascade-connected delay elements to reduce the number of calculations. , and has little interband crosstalk, and is used as the digital filter shown in FIGS. 1 and 2.

When the transfer function of the above half-band filter is G(e ^j 〓 ^T ) (ω=angular frequency, T=sampling period), this antisymmetric filter can be constructed by , by replacing N cascaded delay elements as described above, the transfer function H(e ^j 〓 ^T ) of the circuit becomes H(e ^j 〓 ^T )=G
(e ^jN 〓 ^T ), it can be realized with the same number of calculations as the above-mentioned half-band filter and N times the number of delay elements.

FIG. 6 is a block circuit diagram illustrating an antisymmetric filter implemented in the above method according to a second embodiment of the invention. In the figure, 13 is a delay element, 14 is a multiplier, and 15 is an adder.

FIG. 7 is an example of a half-band filter designed to implement the antisymmetric filter of FIG. One delay element 13a in FIG. 7 is replaced in FIG. 6 by seven delay elements 13 between two adjacent multipliers.

FIG. 8 shows the wave characteristics of the half-band filter shown in FIG.

FIG. 9 shows the wave characteristics of the antisymmetric filter of FIG. As can be seen from the figure, the period of the passband in FIG. 9 is 1/7 of that in FIG. 8.

FIGS. 10a to 10d show the first example when antisymmetric filters having the wave characteristics shown in FIG. 9 are used in place of the digital filters 3 and 10 shown in FIGS. 1 and 2.
The spectrum shown in each part of the circuit shown in the figure is easily inferred based on the explanation of FIG. 5, so a detailed explanation will be omitted.

Furthermore, from the experimental results, as shown in FIG. 6, when the number of delay elements between adjacent multipliers is set to 7, the deterioration of the voice is the least.

Furthermore, in both the wave characteristics shown in Figure 3 and the wave characteristics shown in Figure 9, a certain frequency band is removed from the original sound, but according to experiments, the auditory sense It is known that the voice sounds almost the same as Uehara's voice.

(7) Effects of the Invention As explained above, according to the present invention, by thinning out the sample values of the digital signal and inserting sample values of zero value, it is possible to improve the sine wave generation circuit and amplitude in audio compression and demodulation systems. Since a modulation circuit is no longer required, the number of parts is greatly reduced and the device cost is reduced. At the same time, the use of an antisymmetric filter with good pass/block characteristics with the same number of calculations as a half-band filter improves the auditory sense. This provides the advantage of being able to transmit in half the audio bandwidth while ensuring good audio quality that is almost the same as the original audio.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing a transmitting section that compresses an input audio signal according to an embodiment of the present invention, FIG. 2 is a block diagram showing a receiving section that demodulates the received signal from the transmitting section of FIG. 1, and FIG. The figures are graphs showing the wave characteristics of the digital filters shown in Figs. 1 and 2, Fig. 4 is a graph for explaining the operation of the thinning circuit shown in Fig. 1, and Figs. A waveform diagram showing the spectrum of each part in FIGS. 1 and 2,
FIG. 6 is a block circuit diagram showing an anti-symmetric filter realized according to the second embodiment of the present invention, and FIG. 7 is an example of a half-band filter designed to realize the anti-symmetric filter of FIG. FIG. 8 is a graph showing the wave characteristics of the half-band filter shown in FIG. 7, FIG. 9 is a graph showing the wave characteristics of the antisymmetric filter shown in FIG. 6, and FIG. FIG. 6 is a waveform chart showing the spectrum of each part in FIG. 1 and FIG. 2 when the antisymmetric filter in FIG. 6 is used as the filter. 1... Transmitter, 2... First AD converter, 3... First
digital filter, 4... sample thinning circuit, 5... first DA converter, 6, 7... transmission line, 8
...Second AD converter, 9...Zero sample insertion circuit,
10...Second digital filter, 11...Second digital filter
DA converter, 12... handset, 13... delay element, 1
4... Multiplier, 15... Adder.

Claims (1)

[Claims] 1. When the highest frequency of the input audio signal is F,
On the transmitting side, the input audio signal is AD-processed at a sampling frequency of 2F.
a first analog-to-digital converter for converting the output signal of the first analog-to-digital converter; Digital filter, by thinning out the output signal of the first digital filter every sample time, a sample is created in which signal components from frequency zero to F/2 and signal components from frequency F/2 to F are mutually folded into an empty frequency range. It is equipped with a thinning circuit and a first digital-to-analog converter that converts the output signal of the sample thinning circuit at a sampling frequency F and transmits it, and on the receiving side, the analog signal transmitted from the transmitting side is converted to a sampling frequency. The second analog to AD convert with F.
a digital converter; a zero sample insertion circuit that inserts one sample value of zero between adjacent sample times of the output signal of the second analog-to-digital converter and outputs the sample value as a sampling frequency of 2F; a second digital filter that has the same characteristics as the first digital filter and waves the output signal of the zero sample insertion circuit; and a second digital filter that converts the output signal of the second digital filter at a sampling frequency of 2F and sends it to the receiver. An audio compression and demodulation method comprising a second digital-to-analog converter for outputting. 2. The first and second digital filters each have N (N is an odd number) delay elements included in a digital low-pass filter or a digital high-pass filter with a sampling frequency of 2F and a cutoff frequency of F/2 connected in series. Claim 1 in which the structure is replaced by a connected delay element.
Audio compression and demodulation methods described in Section. 3. The audio compression and demodulation system according to claim 2, wherein the number N of the delay elements is seven.