JP3219892B2 - Real-time speech speed converter - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention
The present invention relates to a real-time speech speed conversion device for converting into a low-speed speech suitable for hearing a voice of a hearing-impaired person or an elderly person.

[Summary of the Invention] [0002] The present invention relates to a real-time speech speed conversion device which takes in an original voice and converts it into a slow-speed voice suitable for listening to a voice of a hearing-impaired person or an elderly person. In the processing, a change in the intonation (pitch frequency) of the voice is detected, and based on the detection result, the speech speed is changed according to the rule that the speech speed is relaxed in the high intonation portion and the speech speed is increased in the low intonation portion. In this way, the original voice is converted into an easy-to-hear good voice while keeping the speech time of the original voice.

[0003]

2. Description of the Related Art Generally, when a listener loses his / her listening ability such as a critical speed for voice recognition (maximum speech speed at which voice can be accurately recognized) due to aging or some kind of obstacle, a normal speed voice or a quick voice is used. The discrimination degree of the spoken voice is greatly reduced.

[0004] Conventionally, there has been only a hearing aid as a hearing aid for a person having such a hearing impairment.

[0005]

The hearing aid described above is a device for compensating only the transmission characteristics of the outer ear and the middle ear of the auditory system by simply improving the frequency characteristics and controlling the gain, so that the hearing aid is not used. There is a problem that it is not possible to compensate for a decrease in the ability to discriminate voices involved in the deterioration.

Therefore, as a method for solving such a problem, a technique for converting the speech speed while maintaining the quality of the original voice has been developed.

In this speech speed conversion technique, it is possible to make the speech much easier to hear, especially for the elderly and the hearing impaired, by uniformly lowering only the speech speed of the speech. The operation also inevitably extends the speech utterance time. However, in broadcasting, etc., the utterance time of the voice before expansion is uttered so as to be within the predetermined time, so if such voice expansion is performed, there is a possibility that the voice will not fit within the above-mentioned time limit. Occurs. Further, in a case where audio and video are provided synchronously as in a television or the like, if only audio is expanded, a temporal "shift" occurs between the audio and the video, which may adversely affect listening. appear.

[0008] For this reason, although a speech speed conversion technique taking into account such a temporal “shift” has been developed by an off-line process, a speech rate conversion technique taking into account the temporal “shift” has been developed. What can be realized by real-time processing has not yet been developed.

The present invention has been made in view of the above circumstances, and in order to solve the above-described problem associated with the time lag, the speech speed of a portion considered to be significant in the uttered voice is appropriately adjusted by real-time processing. It is an object of the present invention to provide a real-time speech speed conversion device capable of converting speech into a slow and easy-to-hear sound as a whole without substantially extending the speech time by slowing down and advancing other portions in reverse. And

[0010]

In order to achieve the above object, according to the present invention, when the speed of uttering a listening sound (speaking speed) is reduced, a silent section for a predetermined time or more is detected. A pause section determining means for determining the silent section as a pause section meaning a breath section of the utterance; a phrase section between the pause section and the next pause section; and a predetermined number from a start point of the phrase section. A maximum pitch frequency detecting means for detecting a maximum pitch frequency in a voiced section, and a predetermined expansion factor from the start point of the phrase section, and based on a predetermined decreasing function. After the lapse of the period, a speech speed conversion means for executing a speech speed conversion in consideration of the maximum pitch frequency is provided. According to a second aspect, in the real-time speech rate conversion device according to the first aspect, the speech rate conversion means determines an average of voiced sections to be processed in the phrase section when determining the speech rate after the lapse of the predetermined time. It is characterized in that a pitch frequency is obtained, and an expansion ratio of a voice section to be processed is determined based on a magnitude relationship between the average pitch frequency and a numerical value obtained by multiplying the maximum pitch frequency by a predetermined threshold.

[0011]

According to the above arrangement, the pause section determining means includes:
When reducing the speed at which the listening sound is uttered (speaking speed), a silent section for a predetermined time or longer is detected, and this silent section is determined as a pause section which means a breath section of the uttered sound. The maximum pitch frequency detecting means sets a phrase section between the pause section and the next pause section, and detects a maximum pitch frequency in a predetermined number of voiced sections from a start point of the phrase section.
Then, the speech speed conversion means, based on a predetermined expansion factor from the start point of the phrase section and a predetermined decreasing function,
The voice speed is converted over a certain period of time, and after a certain period of time, the voice speed conversion is performed in consideration of the maximum pitch frequency. As a result, while solving the problems associated with the temporal "shift", the speech speed of the semantically important part in the uttered speech is moderately reduced, and the other parts are reversed, thereby shortening the utterance time. It converts the sound into slow, easy-to-hear sound in real time, without any substantial stretching.

[0012]

FIG. 1 is a block diagram showing an embodiment of a real-time speech speed converter according to the present invention.

The real-time speech speed converter shown in FIG. 1 includes a voice input circuit 1, a CPU circuit 2, a PROM circuit 3, an input buffer circuit 4, a processing buffer circuit 5,
The voice input circuit 1 includes a file circuit 6, a voice output circuit 7, and a bus 8. The voice input circuit 1 captures voice (original voice) to be converted in speech speed, and performs real-time processing into the original sound intonation (pitch frequency). Of the original voice, and based on the detection result, the speech speed is changed according to the rule that the speech speed is slowed down in the portion of high intonation and the speech speed is increased in the portion of low intonation, thereby keeping the speech time of the original voice. The original sound is converted into a good sound that is easy to hear while keeping it.

The audio input circuit 1 is a circuit having a general configuration for inputting an original audio, for example, a microphone, a tone circuit, an analog-to-digital converter, a voice storage / reproduction (recording).
Circuit, an audio storage medium (for example, an IC memory, a hard disk, a floppy disk or a VTR), an interface circuit, and the like. The audio to be converted is converted into a digital audio signal. The data is supplied to the input buffer circuit 4 in frame units based on an instruction from the CPU circuit 2.

The input buffer circuit 4 has a required capacity RA.
M, and is used as a work area of the CPU circuit 2. The audio signal output from the audio input circuit 1 is taken in and stored therein.
The stored audio signal is transferred to the processing buffer circuit 5 based on an instruction from the circuit 2.

The processing buffer circuit 5 has a required capacity RA.
M, and is used as a work area of the CPU circuit 2. The audio signal output from the input buffer circuit 4 is fetched and stored.
The stored audio signal is transferred to the file circuit 6 or the like based on an instruction from the PU circuit 2.

The file circuit 6 is constituted by an audio storage medium such as an IC memory or a floppy disk, in addition to the RAM, and performs the processing of expanding the voiced section of the voiced section and shortening the silent section according to the present invention. A memory for storing a processed audio signal when the processed audio signal is output from the processing buffer circuit 5 and storing the same, and thereafter storing the audio signal based on an instruction from the CPU circuit 2. Is supplied to the audio output circuit 7.

The audio output circuit 7 is a circuit having a general configuration for outputting the audio signal in the file circuit 6 to the outside, for example, an interface circuit, a digital-to-analog converter,
A speaker, a recording device (or a broadcasting device), etc., and when an audio signal is output from the file circuit 6,
It captures this and converts it into audio while outputting it to the outside.

The CPU circuit 2 is a portion constituted by a one-chip microcomputer or the like.
Based on a program stored in the OM circuit 3, the control of the entire apparatus and various data processing are performed.

The PROM circuit 3 is a portion used as a storage location for programs that define the operation of the CPU circuit 2 and constant data used in various processes.
In accordance with a read command from the CPU circuit 2, the stored program or constant data is read and supplied to the CPU circuit 2.

<< Operation of Embodiment >> Next, the operation of this embodiment will be described with reference to the block diagram shown in FIG. 1, the flowcharts shown in FIGS. 2 and 3, and the timing chart shown in FIG.

First, the CPU circuit 2 cuts out an audio signal which is first input to the audio input circuit 1 and processed and is cut out at a fixed length called a frame, for example, every 3.3 ms, transferred to the input buffer circuit 4 and stored therein. (Step ST
1).

Thereafter, the CPU circuit 2 processes the audio signal stored in the input buffer circuit 4 for each frame by a method such as the auto-correlation method or the zero-cross method, and performs voiced, unvoiced, and silent operation for each frame. Make a decision. However, in this case, input sounds other than voiced and unvoiced sounds (for example, low-level noise and background sounds) generated by a person are processed as silence in principle (step ST2).

Next, the CPU circuit 2 determines whether or not the determination result of voiced, unvoiced, and silent for the current frame is the same as the determination result of voiced, unvoiced, and silent for the previous frame (step ST3). If these are the same type, the process returns to the above-described frame extraction processing and the same processing is repeated. If the type is a different type, for example, the previous frame is a voiced section and the current frame is an unvoiced section, until then, The audio signals determined to be the same type of section are transferred to the processing buffer circuit 5 and stored (step ST4).

As a result, as shown in FIG. 4, the voice fetched by the voice input circuit 1 is divided into a voiced section, an unvoiced section and a silent section and stored in the processing buffer circuit 5.

Thereafter, the CPU circuit 2 sets a pause section (sending section) of a section having a section length of 250 ms or more among sections determined as silent sections among the audio signals stored in the processing buffer circuit 5. It is determined that the section is a breath section of a vocal sound) and the section between the pause sections is a phrase section (a section uttered in a single breath) (step ST).
5).

Then, the CPU circuit 2 performs a voiced section process 10 shown in FIG. 3 on the section determined as a voiced section of each phrase section (step ST6).

In the voiced section processing, the CPU circuit 2 first determines whether or not the voiced section to be processed is a voiced section immediately after the pause section (step ST15). , Three voiced sections from the start point (Ph_st) of the phrase section (first voiced section,
The second voiced section and the third voiced section) are extracted, the highest pitch frequency among the pitch frequencies of the first voiced section to the third voiced section is set as the highest pitch frequency Pitch_max, and the start of the first voiced section is started. The expansion rate of the speech speed at the point V_st is set to “rs” (step ST1).
6).

Thereafter, the CPU circuit 2 determines that the audio signal to be processed has a predetermined time T (2000 m in this embodiment) from the start point V_st of the first voiced section.
s) is determined (step ST1).
7) If the time T has not elapsed, the expansion rate of the speech speed is changed from "rs" to "re" by using an appropriate decreasing function set in advance, for example, a cosine function f (t) shown in the following equation. It is changed (step ST18).

F (t) = re + (1/2) · (rs−re) · {cosπ · (t−V_st) /T+1.0} (1) where t: t = V_st to V_st + T At this time, in this range, no processing is performed on the silent section and the unvoiced section.

If the audio signal to be processed has passed a predetermined length of time T from the start point V_st of the first voiced section (step ST17), the CPU circuit 2 becomes the processing target. Section containing the audio signal (the nth
It is determined whether or not the average pitch frequency Pitch (n) in the voice section (where n ≧ k) satisfies the following equation (step ST19).

Pitch (n)> Pitch_max × Th2 (2) where Th2 is a threshold value, and in this embodiment, Th is
2 = 0.7.

If the n-th voice section satisfies the above equation (2), the CPU circuit 2 sets the start point of the n-th voice section to "V2_st" (step ST20) and sets the expansion ratio to "rs-Th3". ". The above-described process of determining whether or not the time period is within the period and the process of expanding the voiced section using the decreasing function f (t) are performed.
The speech speed expansion factor is changed from "rs-Th3" to "re" within the range (step ST18).

In this case, in this embodiment, the threshold value Th
3 is set to “0.1”.

If the expression (2) is not satisfied (step ST19), the CPU circuit 2 keeps the expansion ratio re of the voiced section, that is, the speech speed at the fastest (step ST21).

Thereafter, every time a voiced section is detected until the next pause section, the CPU circuit 2 repeats the above-described processing on the audio signal in this voiced section.

Then, after this processing is completed, the CPU circuit 2 transfers the speech speed-converted audio signal in the processing buffer to the file circuit 6 for storage, and clears the processing buffer circuit 5 (step ST7). ).

In the above-described identification section processing (step ST5), if the section to be processed is determined to be an unvoiced section, the CPU circuit 2 transfers the audio signal in this section from the processing buffer circuit 5 to the file circuit 6. After that, the processing buffer circuit 5 is cleared (step ST
8).

In the above-described identification section processing (step ST5), if the section to be processed is determined to be a silent section, the CPU circuit 2 determines whether or not this section is a pause section (step ST9). In the case of, it is determined that the sentence is a break (punctuation point), and in order to shorten the sentence to the shortest without any discomfort in the sense of hearing, a shortening process of a preset algorithm is performed to shorten a silent section ( Step ST10).

Further, in the above-described identification section processing (step ST5), if it is determined that the section is a silent section but is not a pause section (step ST9), the CPU circuit 2 skips this shortening processing.

Thereafter, the CPU circuit 2 sends the processed silence section signal in the processing buffer circuit 5 to the file circuit 6.
After that, the processing buffer circuit 5 is cleared (step ST11).

Thereafter, the CPU circuit 2 repeats the above processing until there is no more audio signal to be processed (step ST12).

In parallel with the above processing, the CPU circuit 2 transfers the processed audio signal stored in the file circuit 6 to the audio output circuit 7 and outputs it as audio.

<< Experimental Example >> When the speech including the actual news speech shown in Table 1 is processed by the above-described real-time speech speed converter, (1) -1 to (1) -6 in the text are Recognizing each break as a pause, the speech speed can be reduced at the start point of each phrase. (1) -1, (1)
In the latter part of -4 (underlined part), the speech speed could be reduced.

[0045]

[Table 1] In this experiment, the expansion rate of the speech speed was set to “rs =
When “1.25” and “re = 0.9”, the original voice having a length of 136 seconds can be converted into a voice having a length of 136 seconds, and “rs = re = 1.3” and the speech speed Compared with the case where the absorption rate is uniformly slowed, the absorption rate α is set to 100% (the non-absorption time 0.0
Seconds).

In this case, the absorption rate α is a value represented by the following equation.

Α = {(T1−T2) / T1} · 100 (3) where T1: the extension time when the speech speed is uniform T2: the extension time when the speech speed is changed Therefore, in this embodiment, By using, the silent section between sentences is effectively shortened, so that the original speech can be converted into a slow speech in real time without extending the entire time length.

As described above, in this embodiment, the voice (original voice) to be subjected to the speech speed conversion is fetched by the voice input circuit 1, and the change of the intonation (pitch frequency) of the original voice is detected by real-time processing. Based on this detection result, in the part with high intonation,
By reducing the speech speed and changing the speech speed in the low part at the rule of increasing the speech speed, the original voice is converted into a good voice that is easy to hear while maintaining the speech time of the original voice. The speech speed of the important parts in the uttered speech can be moderately slowed, and the other parts can be sped up in reverse, so that it is easy to listen as a whole without substantially increasing the speech time. Can be converted to voice.

[0049]

As described above, according to the present invention, in order to solve the problem associated with the temporal "shift", the real-time processing is performed on the part of the uttered speech that is considered to be a significant part. By slowing the speed appropriately and increasing the other parts in reverse, it is possible to convert the speech into a slow, easy-to-hear sound as a whole without substantially extending the speaking time.

[Brief description of the drawings]

FIG. 1 is a block diagram showing one embodiment of a real-time speech speed conversion device according to the present invention.

FIG. 2 is a main flowchart showing an operation example of the real-time speech speed conversion device shown in FIG.

FIG. 3 is a flowchart showing an example of a voiced section processing routine showing an operation example of the real-time speech speed conversion device shown in FIG. 1;

FIG. 4 is a timing chart showing an operation example of the real-time speech speed conversion device shown in FIG. 1;

[Explanation of symbols]

 Reference Signs List 1 audio input circuit 2 CPU circuit 3 PROM circuit 4 input buffer circuit 5 processing buffer circuit 6 file circuit 7 audio output circuit 8 bus

──────────────────────────────────────────────────続 き Continued on the front page (58) Field surveyed (Int.Cl. ⁷ , DB name) G10L 21/04

Claims (2)

(57) [Claims] 1. A pause for detecting a silent period of a predetermined time or more when the speed of uttering a listening voice (speaking speed) is reduced, and determining the silent period as a pause period representing a breath interval of the uttered sound. Section determination means; a phrase section between a pause section and the next pause section; a maximum pitch frequency detection section for detecting a maximum pitch frequency in a predetermined number of voiced sections from a start point of the phrase section; Based on a predetermined expansion factor and a predetermined decreasing function from the start point, the speech speed is converted over a certain period of time, and after the lapse of a certain period of time, the speech speed is converted in consideration of the maximum pitch frequency. A real-time speech speed conversion device, comprising: conversion means. 2. The real-time speech speed conversion device according to claim 1, wherein the speech speed conversion means determines an average of a voiced section to be processed in the phrase section when determining the speech speed after the predetermined time has elapsed. A real-time speech speed conversion apparatus characterized in that a pitch frequency is obtained, and an expansion ratio of a voice section to be processed is determined based on a magnitude relationship between the average pitch frequency and a numerical value obtained by multiplying the maximum pitch frequency by a predetermined threshold.