JPS6323425A - Sound reproducing system - Google Patents

Abstract

PURPOSE:To improve the sound quality of a reproduced sound by superimposing a prescribed noise signal for a period having a prescribed time or below of the duration time among periods discriminated to be silence periods. CONSTITUTION:A decoding expansion section 11 identifies a silence flag FN at the head of an inputted coded data DC and supplies it to a silence period control section 12, which checks whether or not its block is consecutive for the number or over corresponding to a prescribed time T1 when the content of the silence flag FN is included in the silence period. If the result is NO, the selection signal SL outputted to a selector 14 is descended as to all blocks discriminated to be silence periods to allow the selector 14 to select a noise data DZ fed to an input terminal A. In the block constituting a silence period having a shorter duration time than a prscribed time T1 among silence periods, the noise data DZ is fed to a digital/analog converter 19 as a sound data SD. Thus, the signal component missing partially is compensated by the superimposed noise signal to improve the sound quality of reproduced sound.

Description

DETAILED DESCRIPTION OF THE INVENTION [Technical Field] The present invention relates to an audio reproduction method for reproducing audio data in which silent sections have been compressed.

[Prior Art] For example, when digitally processing a voice signal, such as storing and synthesizing voice signals for a voice response device, it is necessary to convert the voice signal into a digital signal by some method.

Basically, audio signals have a frequency band of 0.3 to 3.4K)
This is an analog signal of Iz, and in order to convert it to a digital signal, it can be converted using an analog/digital converter with a sampling frequency of 8 KHz and a resolution of 8 bits (PCM (Pulse Cods Modulat)).
ion) encoding).

Then, to convert this digital signal back to the original audio signal,
It is sufficient to convert it into an analog signal using a digital/analog converter with a sampling frequency of 8 KHz and a resolution of 8 bits, and then pass it through a low-pass filter to shape the waveform. The larger the bit width of the code, the higher the quality of the reproduced audio.

By the way, such a PCM encoded audio signal has 1
The bit rate (data rate; hereinafter referred to as bit rate) per second is 64 bps, and in order to store audio signals with such a high bit rate, a memory with a huge storage capacity is required. Therefore, various proposals have been made to reduce the bit rate of audio signals.

One of these proposals is to reduce the amount of information in the overall audio data by encoding the so-called silent sections, where the amplitude is very small between words or within words, in the audio signal. There is a so-called silence compression method that does this.

In such a silence compression method, the amount of information in audio data is reduced to 2
It can be reduced by about 0 to 25%.

Now, when an audio signal is reproduced from audio data whose information amount has been reduced by such a silence compression method, conventionally, the audio signal is reproduced with an amplitude of O in an interval determined to be a silent interval.

However, in such a reproduction method, the sound including the silent section is reproduced as a sluggish sound that lacks smoothness, resulting in a problem of poor sound quality.

[Purpose] The present invention has been made to solve the above-mentioned disadvantages of the prior art, and an object of the present invention is to provide an audio reproduction method that can improve the sound quality of reproduced sound.

[Configuration] In order to achieve this object, the present invention superimposes a predetermined noise signal on a section whose duration is equal to or less than a predetermined time among sections that are determined to be silent sections.

ing.

Hereinafter, one embodiment of the present invention will be described in detail with reference to the accompanying drawings.

First, the principle of the present invention will be explained.

For example, when considering two words that are pronounced consecutively in one voice message data, the speech waveforms of the two words will have a gap between the words as shown in Figure 1(a) when viewed from the envelope. The amplitude of the part becomes very small,
takes the value "0".

Also, considering the speech waveform within the same word, Fig. 2(a)
As shown in Figure 2, the sounds that are produced include so-called unvoiced sounds whose amplitude is very small.

In other words, for the silent sections and unvoiced parts between words, the duration of the silent section is more important than the amplitude information of the speech waveform, and if such duration information can be encoded, the speech signal can be encoded. Can be played almost properly.

Therefore, in order to determine this silent section, for example, an amplitude a that is a threshold value is set, and if a portion with an amplitude smaller than this amplitude a is determined as a silent section, in Fig. 1 (a), the sections All , A12. A13 is determined as a silent section.

If these silent sections are played back without any sound, an audio signal with a waveform as shown in Figure 3(b) will be reproduced, but the sound signal that was included in the original audio signal in these silent sections will be played back. Because the small-amplitude signal is missing, the reproduced sound is perceived as unnatural audibly.

Therefore, in order to eliminate such audible unnaturalness,
First, it may be possible to compensate for small-amplitude signal components that are missing in silent sections. To do this, a predetermined noise signal of small amplitude may be superimposed on the silent section of the reproduced audio signal.

However, if a noise signal is superimposed over the entire period of all silent sections, the S/N ratio of the reproduced audio signal will deteriorate, so it is necessary to appropriately set the portion on which the noise signal is superimposed.

Here, if we examine the audio signal in Figure 1(a), we can see that in the section that is determined to be a silent section, a small-amplitude signal is added to the one-dimensional audio signal in the part that is in contact with the non-silent section. ,
It is effective to superimpose a noise signal on this part, but
The central portion of a long silent section is often a true silent state, and therefore it is better not to superimpose a noise signal on the long silent section. Or, in long silent sections,
It is better to superimpose a noise signal at the edges and not to superimpose a noise signal at the center.

According to this method, as shown in FIG. 1(C), the noise signal NZ is superimposed and the unnaturalness of the reproduced sound is eliminated.

Furthermore, as shown in Fig. 2(b), the silent section within a word is very short, so it is not effective to superimpose the noise signal NZ on such a short silent section as shown in Fig. 2(c). be.

FIG. 3 shows a speech encoding device according to an embodiment of the present invention. In this audio encoding device, a sample of an audio signal is divided into a predetermined number of blocks, and silent periods are identified in units of blocks.

In the same figure, input audio data SS is band-limited by a low-pass filter 1 and set to a predetermined sampling frequency (for example, 8
Kllz) is converted into a digital signal DS (PCM code) with a predetermined number of bits (e.g. 8 bits), and this digital signal O8 is applied to the encoding compression section 4 via the delay memory 3, and is also applied to the silent section detection section 5. has been added.

When the silent section detection unit 5 receives the digital signal DS of the number of samples for one block, it determines whether or not the block is a silent section using a well-known method, and if it determines that it is a silent section, it detects silence. The signal DN is output to the encoding/compressing section 4. As a method for determining a silent section, for example, when the maximum amplitude in one block is smaller than a predetermined value, there is a method of determining the block as a silent section.

The encoding compression section 4 divides the digital signal DS applied via the delay memory 3 into blocks each having a predetermined number of samples, and for blocks for which the silence detection signal DN is not output from the silence section detection section 5, A block data DB formed by encoding the digital signal DS of a block using a well-known method, and a silence flag that is added to the beginning of this block data DB and set to 1 to indicate that the block is a non-silence section. For blocks in which coded data DC (see FIG. 4(a)) consisting of FN is formed and a silence detection signal DN is output, the content indicating that the block is a silent section is set to 1. Only the silence flag FN is formed as coded data DC, and this coded data DC is output to the next stage device.

Therefore, as shown in FIG. 4(b), the encoded data DC output from the encoder and compressor 4 has a silence flag F whose contents are set to 1 for blocks determined as silent sections.
Since it consists of only N, the amount of information of the audio data is reduced.

In this case, block n+1. n+2°n+5 is determined to be a silent section.

The encoding and compression processing carried out by the encoding and compression unit 4 includes, for example, a differential PCM encoding method that forms a difference between chronologically adjacent PCM codes, and an adaptive differential PCM that is a further advancement of this differential PCM encoding method. Encoding method (e.g. CC
ADP as recommended by ITT (International Telegraph and Telephone Advisory Committee)
CMf), APC-AB (Adaptive Predic
tionCoding with Adaptive
Bit A11ocation) method, LSP (Line 5pectruIIIPa) using speech analysis and synthesis method
ir) method, or the quasi-instantaneous companding method, which is adopted as one of the high-quality PCM audio transmission methods for broadcasting satellites. By such encoding compression, the bit rate of audio data is reduced to, for example, 32 Kbps or less.

FIG. 5 shows an example of an audio decoding device according to an embodiment of the present invention. This audio decoding device decodes the encoded data DC described above.

In the figure, the decoding and decompression unit 11 identifies the silence flag FN at the beginning of the input encoded data DC and outputs it to the silence section control unit 12, and when the content of the silence flag FN is O, the following The block data DB is decoded by a predetermined method and the decoded data CD is outputted to the input end B of the selector 14 via the buffer memory 13. Further, when the content of the silence flag FN is 1, decoded data CD is output in which all samples in the block have an amplitude of 0.

Further, the other input terminal A of the selector 14 is connected to an oscillator 15 that generates a pulse signal PC at a sampling frequency, a counter 16 that counts the pulse signal PC, and a ROM (read-only) whose address is set according to the count value of the counter 16. Noise data DZ output from a noise generator 18 consisting of a memory) 17 is added.

As shown in FIG. 6, the silent section control unit 12 identifies whether the content of the silent flag FN is O and determines whether or not the block is included in the silent section.
1), when the result of this judgment 101 is YES, it is checked whether the block is consecutive for a predetermined time T1 or more (judgment 102), and when the result of this judgment 102 is NO, the silence is For all blocks determined as sections, the selection signal SL output to the selector 14 is raised to cause the selector 14 to select the noise data DZ applied to the input terminal A (process 103).
, for other blocks, the selection signal SL is lowered to cause the selector 14 to select the decoded data CD applied to the input terminal B.

As a result, the selector 14 outputs noise data DZ in blocks that constitute a silent section whose duration is shorter than the predetermined time T1.

For blocks constituting other silent sections and blocks constituting non-silent sections, decoded data C
D are respectively applied to the digital/analog converter 19 as audio data SD. Here, the predetermined time T1 is set to about 10 milliseconds.

According to the above-mentioned sampling conditions, one block constitutes one millisecond, so the number of blocks corresponding to this predetermined time T1 is ten.

The digital/analog converter 19 converts the input audio data SD into a corresponding analog signal (level signal) at a predetermined conversion frequency, and this analog signal is waveform-shaped by a low-pass filter 20 and is then reproduced as a reproduced audio signal. output to the stage device.

Therefore, in a non-silent section, an audio signal corresponding to the decoded data CD is used, in a silent section shorter than a predetermined time T1, an audio signal corresponding to the noise data DZ is used, and in other silent sections, a silent audio signal is used. As a result, an audio signal with noise superimposed in a silent section shorter than the predetermined time T1 is reproduced.

Another example of the process executed by the silent section control unit 12 is shown in FIG.

In this process, for a silent section lasting longer than a predetermined time T1, noise data DZ is superimposed on the first predetermined time T2 and the last predetermined time T3 of the silent section. Furthermore, for silent sections whose duration is shorter than the predetermined time T1, the noise data DZ is superimposed on all of the silent sections. Here, the predetermined time T2 is set to 5 milliseconds, and the predetermined time T3 is set to 10 milliseconds, for example.

As a result, an audio signal corresponding to the decoded data CD is played in a non-silent section, an audio signal corresponding to the noise data DZ is played in a silent section shorter than a predetermined time T1, and other silent sections are played back. In the section,
The first predetermined time T2 and the last predetermined time T of the silent section
An audio signal corresponding to the noise data DZ is reproduced in the portion 3, and a silent audio signal is reproduced in the other portions.

As a result, an audio signal is reproduced with noise superimposed at both ends of a silent section shorter than the predetermined time T1 and a silent section longer than the predetermined time T1.

Further, FIGS. 8 and 9 show still other examples of processing executed by the silent section control unit 12.

In the process shown in FIG. 8, in a silent section longer than a predetermined time T1, the noise data DZ is generated at the first predetermined time T2.
is selected, and noise data DZ is selected for all silent sections shorter than the predetermined time T1. As a result, an audio signal with noise superimposed at the start end of a silent section shorter than the predetermined time T1 and a longer silent section is reproduced.

In the process shown in FIG. 9, in a silent section longer than a predetermined time T1, noise data DZ is generated at the last predetermined time T3.
is selected, and noise data DZ is selected for all silent sections shorter than the predetermined time T1. As a result, an audio signal with noise superimposed at the end of a silent section shorter than the predetermined time T1 and a longer silent section is reproduced.

By the way, the noise generator 18 generates the noise data DZ by scanning a small amplitude noise waveform (for example, white noise) stored in the ROM 17 using the count value of the counter 16, but the storage capacity of the ROM 17 is limited. Since it is finite, the noise data DZ has periodicity. Since it is not preferable that the periodicity of the noise data DZ is audibly sensed, the sample data stored in the ROM 17 is set so that the period of the noise data DZ is below the human audible frequency.

That is, the human audible range is said to be approximately 20 Hz to 20 Hz. Therefore, the period of noise data DZ is set to 20.
If the value is smaller than Hz, the periodicity of the noise data DZ will not be audibly perceived.

Therefore, if the sampling frequency is set to 8 Hz,
By configuring the noise data DZ with a number of samples greater than 400 (=8000/20), generation of such periodic noise can be prevented.

FIG. 10 shows another example of the audio decoding device. In addition, in this figure, the same parts as in FIG. 5 are given the same reference numerals, and the explanation thereof will be omitted.

In the figure, a noise generating section 21 generates a small amplitude analog noise signal AZ, and the analog noise signal AZ is applied to an analog switch 22.

Also, the decoded data CD output from the decoding/expanding section 11
is converted into a corresponding analog signal by a digital/analog converter 19, and then waveform-shaped by a low-pass filter 20 and applied to an analog switch 23. Further, the selection signal SL outputted from the silent section control section 12 is applied to the control input terminal of the analog switch 22 and is also applied to the control input terminal of the analog switch 23 via the inverter 24.

Therefore, during the period in which the selection signal SL output from the silent section control section 12 is falling, the analog switch 22 is turned off at the same time as the analog switch 23 is turned on, so that an audio signal corresponding to the decoded data CD is output. Furthermore, during the period in which the selection signal SL is rising, the analog switch 22 is turned on and the analog switch 23 is turned off at the same time, so that the analog noise signal AZ is output.

Similar to the embodiment described above, an audio signal with noise superimposed thereon is reproduced.

Note that the signal format of encoded data and the configuration of the audio encoding device are not limited to those described above. Furthermore, the predetermined times Tl, T2 . The setting time of T3 is not limited to that described above either.

[Effect] As explained above, according to the present invention, a predetermined noise signal is superimposed on a section whose duration is less than a predetermined time among sections that are determined to be silent sections, and this noise signal causes silence. Since the small amplitude signal components that are partially missing in the section are compensated for, the sound quality of the reproduced sound can be improved.

[Brief explanation of the drawing]

1 and 2 are waveform diagrams for explaining the present invention in detail, FIG. 3 is a block diagram illustrating a speech encoding device according to an embodiment of the present invention, and FIG. 4(a). (b) is a signal arrangement diagram showing an example of encoded data. FIG. 5 is a block diagram showing an example of an audio decoding device according to an embodiment of the present invention, FIG. 6 is a flowchart showing an example of processing of the silent section control section, and FIG. 7 is a block diagram showing an example of the processing of the silent section control section and other parts. FIG. 8 is a flowchart showing another processing example of the silent section control section, FIG. 9 is a flowchart showing still another processing example of the silent section control section, and FIG. 10 is a flowchart showing still another processing example of the silent section control section. FIG. 3 is a block diagram showing another example of the audio decoding device. 11...Decoding and expansion unit, 12...Silent interval control unit,
13...Buffer memory, 14...Selector, 15
...Oscillator, 16...Counter, 17...ROM
(Read-only memo January, 18.21... Noise generator, 22.23... Analog switch, 24.
...Inverter. Agent Patent Attorney Crest 1) Makoto' Figure 1 Figure 2 Figure 3 (a) (b) Figure 5 Figure 6 Figure 7 Figure 8 Figure 9 Figure 10

Claims (2)

[Claims] (1) In an audio reproduction method for reproducing audio data in which a silent section is compressed, if the section determined to be a silent section does not continue for a predetermined period of time or more, a predetermined noise signal is superimposed on the silent section. Audio playback method. (2) In an audio reproduction method that reproduces audio data in which a silent section is compressed, if the section determined as the silent section does not continue for a predetermined period of time or more, a predetermined noise signal is superimposed on the silent section, and the silent section is When a section that is determined to be a section continues for a predetermined time or more, the above-mentioned noise is generated during at least one period from the start of the silent section to the end of the predetermined time or from the end to the start of the predetermined time. An audio reproduction method characterized by superimposing signals.