JP2617851B2 - Acoustic sorting introduction method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a case where a specified desired sound is included in an external sound propagating outside a space surrounded by a sound insulating boundary surface (hereinafter referred to as a sound insulating space). Regarding the method of introducing into the space, in particular, it is characterized in that whether or not a specific desired sound is present in the external sound is instantaneously determined by using a computer, and the on / off control of the electrical introduction circuit of the external sound is characterized in that The purpose of the present invention is to provide a method for introducing and sorting sound.

[0002]

2. Description of the Related Art In recent years, the sound insulation material and the sound insulation structure have been improved, so that the sound insulation at the boundary surface has been remarkably improved, and a sound insulation space in which external sound hardly enters can be easily obtained. More people work and enjoy music. However, external sound information that is originally required is not transmitted to a person in such a sound-insulating space, causing various inconveniences. For example, there is a problem in that a person enjoying music in a soundproof room or a person riding a car cannot transmit an emergency warning sound emitted from a firefighter or a police car, and thus cannot sufficiently avoid danger. Conventionally, as means for transmitting external sound into the sound insulation space, there is a telephone means such as an intercom, but external operation is required, which is inappropriate for solving the above-mentioned problems.

[0003]

When an intercom is used as a means for selectively introducing an external sound into the sound insulation space, it is necessary to judge the sound by an external operator, and the desired sound generated at an unintended time is always determined. It is difficult to be able to cope with it, it is easy to cause delay in operation, and the above-mentioned hearing sound judgment is unstable due to individual differences of external operators. Was impossible to maintain.

[0004]

According to the present invention, a computer compares a foreign sound with a desired sound, and automatically introduces only the foreign sound including the desired sound into the sound insulation space according to the degree of approximation. This solves the problem of the conventional means using an intercom. That is, in the present invention, the desired sound waveform is stored in advance in the signal processing device of the computer, and this is compared with the input foreign sound waveform.
A sound selection and introduction method characterized in that on and off control of the extraneous sound reproduction circuit is performed according to the degree of approximation to introduce extraneous sound into the sound insulation space.

Conventionally, various methods for calculating the degree of approximation of two voices using a computer have been proposed as voice recognition means. It has not yet been proposed to use it for screening. The present invention can use the above-mentioned known computer-based voice comparison means, but the conventional voice comparison means is 1) the degree of approximation of the data transition pattern on the time axis 2) the approximation of the transition of the correlation coefficient on the time axis Degree 3) Distance calculation between coefficient data by linear prediction analysis and other methods are available, but these methods use the speech waveform of a minute section as data as it is, so the amount of computer input data is enormous and the calculation becomes complicated. However, when the feature is considered as a sound waveform of a longer time as in the present invention, a data shift is likely to occur and identification tends to be difficult. If a large computer is required and the S / N (signal-to-noise ratio) is low, avoid the adverse effects of background noise on the identification of acoustic waveforms in minute sections There were problems such as not being able to do.

In order to solve the above-mentioned problems in the conventional speech recognition means, the present invention reduces the amount of acoustic information and estimates the power density spectrum of sound which can clarify the characteristics of the acoustic waveform. By using a selection means that can easily and accurately compare and determine a desired acoustic waveform stored in advance with a small-capacity personal computer and an inputted external acoustic waveform by using a small-capacity personal computer, This is a method that is automatically introduced into the sound insulation space. In addition, the power density spectrum of the above sound is the average of the periodogram, and in practice, it is possible to use the value that is reduced in variance and simultaneously smoothed by the method of Boerut,
It is desirable in order to prevent variation in calculation and facilitate calculation.

[0007] As the sound comparison and discrimination means using the estimated amount of the power density spectrum of sound, mainly the following method is appropriate. 1) Correlation coefficient between the estimated amount of the power density spectrum of the desired sound and the extraneous sound 2) Residual energy ratio of the estimated amount of the power density spectrum of the desired sound and the extraneous sound 3) From the characteristic frequency of the desired sound Using the obtained filter, and using the estimated amount of the power density spectrum as the extraneous sound, the two sounds are compared and sorted based on either the energy ratio of the latter passing through the former filter. The above three methods use the desired sound and noise ratio (S /
When N) is large, or when the data structure of the desired acoustic waveform is simple, good selection can be made by using each of them independently, but the S / N is small and the extraneous sound is generated during the extraneous sound. If noise of the same kind of waveform similar to the waveform is included, the possibility of erroneous determination increases.

In the case of an external sound having a bad condition as described above, by combining the above-mentioned comparison determination means 1) and 2), especially in the case of a broad frequency characteristic and a small fluctuation of the external desired sound waveform, Identification can be performed advantageously,
Alternatively, by combining the above-described comparison determination means 1) and 3), the identification can be advantageously performed particularly in the case where the sound waveform has a pure tone characteristic and has an excellent specific frequency distributed in a certain band. Since both of the above 2) and 3) are means for comparing and detecting both acoustic energy ratios, even if they are combined, improvement in the sorting performance cannot be obtained.

[0009]

According to the present invention, it is possible to continuously and automatically determine whether or not a desired sound is included in a foreign sound by the computer according to the above-described structure, and to introduce only the foreign sound having the desired sound into the sound insulation space. In this case, the estimated amount of the acoustic power density spectrum is used as the acoustic information, so that it is possible to easily and reliably select and introduce even a small-sized personal computer.

[0010]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a basic layout of an apparatus for carrying out the method of the present invention, wherein 1 is a microphone for receiving an external sound, 2 is an input amplifier, 3 is a signal processing apparatus by a computer, and a low-pass filter 31, A / D converter 32,
It comprises a central processing unit 33, a D / A converter 34, and a low-pass filter 35. Reference numeral 4 denotes an output amplifier, and reference numeral 5 denotes a speaker. The components other than the microphone 1 are provided in a sound insulation space D surrounded by a dashed line. FIG. 2 is a flowchart showing a method for calculating the power density spectrum estimation amounts S (ω) and A (ω) for the desired sound S and the extraneous sound A in the signal processing device 3, where F is a frequency analysis circuit. .

Example 1 This example shows a method for selectively introducing siren sounds of an ambulance, a police car, and a fire engine as relatively similar sounds. First, sampling frequency 32kHz, F
The average of 200 times of the 50% overlap of the FT512 points (measurement time: 1.6 seconds) was calculated by the calculation in FIG. 2 to obtain the estimated power density spectrum of each siren sound.
FIGS. 3, 4 and 5 are graphs respectively showing the distribution of the estimated amount (vertical axis) of the power density spectrum with respect to the frequency (horizontal axis) of the siren sound of an ambulance, a police car, and a fire engine. Is shown.

FIG. 6 shows a power density obtained by storing any one or all or all of the estimated amount S (ω) of the power density spectrum as a desired sound value in a computer and calculating the extraneous sound in the same manner as the desired sound. Estimate A of spectrum
(Ω) is calculated, and when the two values are compared with each other and the value is equal to or greater than a predetermined threshold value, the sound introduction switch is turned on, the extraneous sound reproduction circuit is closed, and the extraneous sound is introduced into the sound insulation space D. It is a flowchart of the selection method which introduces a sound. The threshold value of the correlation coefficient is determined according to the variation of the acoustic waveform due to the individual difference of the extraneous sound source. If the threshold value C is set to 0.7 based on the experimental result of the siren sound of an actual ambulance or the like, The external sound including the siren sound was introduced without being dropped.

Example 2 In this example, both the comparison and discrimination between the power density spectrum estimation amount S (ω) of the desired sound and the estimation amount A (ω) of the external sound power density spectrum stored in the computer in the case of Example 1 are performed. The correlation coefficient γ of
The residual energy of the A (ω) portion extending outside the pattern of (ω) is determined by the ratio of the total energy of the pattern of A (ω) to the degree of approximation of both acoustic waveforms, and FIG. It is a flowchart of a sorting method. According to the residual energy ratio method, when the target sound is selected and introduced in the same manner as in Example 1, the threshold d of the residual energy ratio is set to 0.2, and the introduction of the external sound below the threshold into the sound insulation space is performed in Example 1. It was possible to carry out in almost the same way as the case.

Example 3 This example uses a filter F (ω) obtained from a characteristic frequency in place of the power density spectrum estimator S (ω) of the desired sound stored in the computer to estimate the power density spectrum of the extraneous sound. The similarity between the two acoustic waveforms is determined based on the ratio of the energy R of A (ω) passing through F (ω) to the total energy of A (ω). Under the same conditions, the threshold value e of the ratio was set to 0.8, and the introduction of the extraneous sound having the threshold value or more into the sound insulation space could be performed in exactly the same manner as in the previous example.
FIG. 8 is a flowchart of the sorting and introducing method.

The correlation coefficient method, the residual energy method, and the filter method shown in the above Examples 1 to 3 each have an individual selection action, have a complicated desired acoustic waveform, and have a high S / N ratio of external sound.
Is low, the same result is not always obtained. In the case where the conditions cannot be met by the respective sorting means as described above, a favorable sorting result can be obtained by combining the correlation coefficient method with another sorting method.

Example 4 As shown in the flowchart of FIG. 9, this example is a selection and introduction method using both the correlation coefficient method and the residual energy method.
Adjusting the respective thresholds c and d is particularly effective when the desired sound has a broad frequency characteristic and the fluctuation of the desired sound waveform is small.

Example 5 This example is a selection and introduction method using both the correlation coefficient method and the filter method as shown in the flowchart of FIG. 10. By adjusting the respective thresholds c and e, especially the S / N
This is effective for the case where the external sound waveform is pure tone and is distributed in a certain frequency band with its outstanding specific frequency.

In the above-described embodiment, the power density spectrum estimation amount S for the desired sound and the external sound is determined.
The discrimination accuracy can be improved by increasing the sampling frequency and the average number of samples for calculating (ω) and A (ω), but the calculation time for frequency analysis also increases with this. In particular, when it is necessary to introduce discrimination in a short time, such as in an emergency alert, even if the discrimination accuracy is reduced to some extent, it is unavoidable. Therefore, it is necessary to select the above various selecting and introducing means and adjust the setting conditions for the comparison and discrimination so as to conform to the required condition of the desired sound in the sound insulation space and the external sound condition.

[0019]

As described above, according to the present invention,
Using a computer to instantly determine whether or not the desired sound is present in the external sound reaching outside the sound insulation space, it is possible to automatically introduce only the foreign sound including the desired sound into the sound insulation space, Therefore, it is possible to prevent danger by selectively listening to emergency warning sounds such as police cars for people in building vehicles where foreign sounds are difficult to reach, or to prevent bird sounds, insect sounds,
The effect of easily and reliably forming a comfortable acoustic space by selecting and introducing a desired sound such as a babbling sound can be obtained particularly by using a small-capacity personal computer.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a basic arrangement of an apparatus for implementing the method of the present invention.

FIG. 2 is a flowchart illustrating a method of calculating an estimated amount of a power density spectrum of sound.

FIG. 3 is a graph showing a distribution of an estimated amount of a power density spectrum of a siren sound of an ambulance.

FIG. 4 is a graph showing an estimated amount of a power density spectrum of a siren sound of a police car.

FIG. 5 is a graph showing a distribution of an estimated amount of a power density spectrum of a siren sound of a fire engine.

FIG. 6 is a flowchart showing a selection method for comparing a desired sound and an extraneous sound by a correlation coefficient method for estimating a power density spectrum of both of them.

FIG. 7 is a flowchart illustrating a selection method for comparing a desired sound and an extraneous sound by the residual energy method for estimating the power density spectrum of both of them.

FIG. 8 is a flowchart of a selection method in which a desired sound and a foreign sound are compared based on a ratio of an estimated amount of a power density spectrum of the foreign sound passing through the filter from a characteristic frequency of the desired sound.

FIG. 9 is a flowchart of a selection method for comparing a desired sound and an extraneous sound in combination with a correlation coefficient method and a residual energy method based on estimators of both power density spectra.

FIG. 10 is a flowchart of a selection method for comparing a desired sound and an extraneous sound by using both a correlation coefficient method and a filter method for estimating the power density spectrum of both.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Microphone 3 Signal processor 5 Speaker D Sound insulation space S (ω) Estimated power density spectrum of desired sound A (ω) Estimated power density spectrum of external sound

Claims (6)

(57) [Claims] A signal processing apparatus of a computer stores a desired sound waveform in advance, compares the sound waveform with an input external sound waveform, and controls on / off of the external sound reproduction circuit according to the degree of approximation of the desired sound waveform. A sound selection and introduction method characterized by introducing an external desired sound into the inside. 2. The method according to claim 1, wherein the estimated amount of each power density spectrum is used as the desired acoustic waveform and the extraneous acoustic waveform, and the comparison is performed by a correlation coefficient between the two. 3. The method according to claim 1, wherein an estimated amount of each power density spectrum is used as the desired acoustic waveform and the extraneous acoustic waveform, and the comparison is performed based on a residual energy ratio between the two. 4. A filter obtained from its characteristic frequency is used as a desired acoustic waveform, and an estimated amount of a power density spectrum is used as an extraneous acoustic waveform. A comparison is made between the two, and the latter estimated amount passes through the former filter. The method according to claim 1, wherein the method is performed based on an energy ratio. 5. The method according to claim 1, wherein the estimated amounts of the respective power density spectra are used as the desired acoustic waveform and the extraneous acoustic waveform, and the comparison is performed by comparing the correlation coefficient between the two and comparing the residual energy ratio between the two. Sound sorting introduction method. 6. An estimator of each power density spectrum is used as a desired acoustic waveform and an extraneous acoustic waveform, and the comparison is made by comparing the correlation coefficient between the two and passing through a filter obtained from the characteristic frequency as the desired acoustic waveform. The method according to claim 1, wherein the method is performed based on an energy ratio of an estimated amount of a power density spectrum of an external sound waveform.