JP2843991B2 - How to convert voiceprint patterns - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Industrial Application Field] The present invention analyzes an engine sound generated from an engine of a vehicle or the like or a sound generated from various other sounding bodies when the engine is operated, and classifies the sounding body. The present invention relates to a method for converting a sound pattern in a device for estimating the tone pattern.

(Prior art)

Conventionally, as a method of analyzing this type of measured sound pattern, a design divider, a so-called ten-point divider, has been used, or Japanese Patent Application No. 61-124195, Japanese Patent Application No. 61-177429, filed by the present applicant earlier. And Japanese Patent Application No. 63-70450.

The method using the design divider matches a measured sound pattern (hereinafter, referred to as “measured sound pattern”) with a previously registered sound pattern record (hereinafter, “registered sound pattern”),
This is a method in which the pattern pitch is measured by a design divider and analyzed by comparing and comparing.

Further, the method disclosed in Japanese Patent Application No. 61-124195 is a method in which a measured sound pattern and a registered sound pattern are simultaneously displayed on the same screen and analyzed by comparing and collating.

Also, the one disclosed in Japanese Patent Application No. 61-177429 displays a measured sound pattern on a display device, and there is a unique sound pattern depending on the engine structure. Therefore, the measured sound pattern is compared with a registered sound pattern. Then, it is a method of specifying and analyzing from similar registered sound patterns.

The method disclosed in Japanese Patent Application No. 63-70450 is a method of comparing, collating and analyzing by registering a registered sound pattern displayed on a transmission type display with a measured sound pattern.

[Problems to be solved by the invention]

However, in the above-mentioned conventional methods, the analysis after displaying the measured sound pattern has had to rely on manual work or visual observation, so that when comparing the measured sound pattern with the registered sound pattern, The problem is that some skill is required for comparison and collation because the individual levels of the pattern pitch and frequency components fluctuate from moment to moment, and there are missing frequency components due to the mixing of unnecessary noise components and fading. was there.

In addition, when the number of registered sound patterns is large, it takes a long time to analyze the measured sound patterns, and it is difficult to analyze the measured sound patterns of the sounds generated by a plurality of sounding bodies. .

The present invention has been made in view of the above points, and eliminates the above-mentioned problems, does not require any skill in comparing and comparing a measured sound pattern with a registered sound pattern, and can be used when the number of registered sound patterns is large or when multiple sound elements are used. It is an object of the present invention to provide a method of converting a sound pattern that facilitates analysis of a measured sound pattern.

[Means for solving the problem]

In order to solve the above-described problems, the present invention measures a sound pattern of a sound generated from various sound elements, and estimates the type of the sound element by comparing and comparing the measured sound pattern with a registered sound pattern registered in advance. In a method of converting a sound pattern of a device, a measured sound pattern is converted by a conversion circuit into a first converted measured sound pattern having a harmonic order versus time characteristic of a fundamental frequency, and the first converted measured sound pattern is integrated by an integrating circuit. Is integrated for each specific time unit range
It is characterized in that it is converted into the next converted measurement sound pattern.

[Action]

By performing the conversion method of the sound pattern as described above, the measured sound pattern of the frequency versus time characteristic is converted by the conversion circuit into a first-order converted measured sound pattern of the high-frequency order versus time characteristic,
Since the first-order converted measurement sound pattern is integrated for each range of the measurement time unit, which is an integrating circuit, and converted to the second-order conversion measurement sound pattern, as will be described later, a high-frequency order pair free from pattern loss and noise is used. A converted measurement sound pattern of the time characteristic is obtained. Accordingly, by making the display format of the registered voice print the same as that of the secondary conversion measurement voice print, the comparison and comparison between the measured voice print and the registered voice print becomes extremely easy.

〔Example〕

 Hereinafter, embodiments of the present invention will be described with reference to the drawings.

FIG. 1 is a block diagram showing a configuration of a voiceprint collation system to which a voiceprint pattern conversion method according to the present invention is applied.
In FIG. 1, an engine or other sounding body (hereinafter referred to as a “sounding body”) 2 mounted on a vehicle or the like 1 (hereinafter referred to as a “sounding body generated sound”) 3 transmits a microphone 5 Can be received at Here, the medium 4 may be any one of a gas, a liquid, and a solid.

The sound 3 generated by the sounding body is converted into an electric signal 6 by a microphone 5, amplified to a required level by an amplifier circuit 7, further subjected to frequency analysis by a frequency analysis circuit 8, and has a frequency-time characteristic as exemplified in FIG. Is obtained. In FIG. 2, the vertical axis represents time t, and the horizontal axis represents frequency f. As shown in FIG. 2, the measurement sound pattern 9 includes factors that hinder the analysis of the measurement sound pattern, such as a temporal variation of the component frequency, a partial loss 10 of the pattern, and noise 11. I have. In addition, the measured sound pattern 9 has a variation in component frequency and a partial lack 10 of the pattern when viewed in time. At each time, these have a harmonic relationship of an integral multiple of the fundamental frequency at that time. The harmonic relationship is always constant regardless of time. Focusing on this, the conversion circuit 12 converts the measured sound pattern 9 into a harmonic order versus time characteristic, thereby obtaining a converted measured sound pattern 13 after the first processing as exemplified in FIG. . In FIG. 3, the vertical axis represents time t, and the horizontal axis represents frequency f.

In the converted and measured sound pattern 13 after the first processing, the harmonic order does not fluctuate with time, but a partial omission 14 and noise 15 of the pattern remain as they are. However, the conversion measurement sound pattern 13 after the first processing is further integrated by the integration processing circuit 16 with the value on the harmonic order axis for each specific time unit range in the time axis direction, thereby obtaining the second A converted measurement sound pattern 17 after processing is obtained. FIG. 4 is a view showing the converted measurement sound pattern 17 after the secondary processing. As shown in FIG.
The converted measurement sound pattern 17 after the second processing is a converted measurement sound pattern in which the partial lack 14 and the noise 15 of the existing pattern are eliminated in FIG.

If the display format of the registered sound pattern 19 registered in the file 18 is set to be the same as the display format of the converted measurement sound pattern 17 after the second processing as illustrated in FIG. A registered voiceprint that is equal to or very similar to the later converted measurement voiceprint 17
19 is retrieved and extracted, and compared and collated by the comparison and collation circuit 20, and the model of the sounding body of the generated sound 3 can be determined from the data of the registered sound pattern 19.

In addition, the automatic processing of the analysis can be performed almost consistently from when the sound 3 generated by the sounding body is received by the microphone 5 to when the model of the sounding body 2 is determined by the comparison / matching circuit 20.

The conversion circuit 12 can extract the fundamental frequency at each time by performing, for example, a Fourier transform process on the measured sound pattern 9.

〔The invention's effect〕

As described above, according to the present invention, the following excellent effects can be obtained.

(1) converting the measured sound pattern of the frequency versus time characteristic into a first-order converted measured sound pattern of the harmonic order versus time characteristic by a conversion circuit;
Since the first-order converted measurement sound pattern is integrated into a second conversion-measurement sound pattern by integrating in a specific time unit range as an integrating circuit, conversion of the harmonic order versus time characteristic free from pattern loss and noise is achieved. The measured sound pattern is obtained, and the comparison and comparison with the registered sound pattern becomes extremely easy.

(2) Further, automatic processing of consistent analysis is possible, including the part that has conventionally had to rely on visual observation.

(3) Further, the automation makes it possible to efficiently compare and refer to an enormous number of registered sound patterns that are extremely difficult to perform by visual processing.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a configuration of a voiceprint matching system to which the voiceprint pattern conversion method according to the present invention is applied, FIG. 2 is a diagram showing an example of a measured voiceprint, and FIG. FIG. 4 shows an example of characteristics after the next processing, FIG.
The figure shows an example of a registered sound pattern. In the figure, 1 ... vehicle etc., 2 ... sound generator, 5 ... microphone, 7 ... amplifier circuit, 8 ... frequency analysis circuit, 12 ... conversion circuit, 16 ... integration processing circuit, 18 ... file , 19 ……
Registered voiceprint, 20 ... Comparison circuit.

──────────────────────────────────────────────────続 き Continued on the front page (58) Fields surveyed (Int. Cl. ⁶ , DB name) G01H 3/00 G01H 3/08 G10L 3/00 531 G10L 3/02 301 G10L 7/08

Claims (1)

(57) [Claims] An apparatus for estimating the type of a sounding body by measuring a sound pattern of a sound generated from various sounding bodies and comparing and collating the measured sound pattern with a registered sound pattern registered in advance. In the pattern conversion method, the measurement sound pattern is converted into a first conversion measurement sound pattern having a harmonic order versus time characteristic of a fundamental frequency by a conversion circuit, and the first conversion measurement sound pattern is converted to a specific time as an integration circuit. A method of converting a sound pattern, wherein the method converts a second converted measurement sound pattern by integrating each unit range.