JP2001165766A - Device for evaluating loudness of unsteady noise - Google Patents

Abstract

PROBLEM TO BE SOLVED: To evaluate the loudness of unsteady noise, to clarify noise problems from an auditory point of view, and to plan and take antinoise measures based on the sense of hearing. SOLUTION: Mechanical noise is measured by a noise measuring device 1, and its measurement data is converted to a digital amount and stored in a data storage device 3. The digital data stored in the data storage device 3 is inputted to a digital filter 4 to pass noise components in a critical band within the audio-frequency region to rectify a pass sound-pressure waveform for every band by a half-wave rectifier 5. The envelope of each rectified waveform is obtained by an envelope computing unit 6 to output a corresponding sound level to a level computing unit 7 to convert levels. Loudness for every critical band is obtained by an auditory characteristic device 8 from the conversion output of the level computing unit 7. A change in loudness which attracts human attention is taken as a characteristic amount by statistical analysis by a characteristics extractor 9, and total loudness ZH is computed from the characteristic amount by a neural network device 10 for unsteady sound and displayed on a display 11.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an apparatus for evaluating loudness of unsteady noise applied to industrial machines such as construction machines and vehicles, and consumer machines such as air conditioners.

[0002]

2. Description of the Related Art Conventional mechanical noise measures are taken based on the loudness and loudness of a noise measured by a noise measuring instrument, and noise measures are taken. By reducing the loudness, the surrounding environment is protected. Have been. Volume is usually "dB"
The standard of evaluation has been determined, and the purpose of noise control was to reduce the noise to or below the level specified by law by measuring the number of decibels (level) of noise.

However, with regard to unsteady noise whose sound pressure fluctuates with time, the dynamic characteristics of the sound level meter do not accurately approximate the dynamic characteristics of the human auditory period, so that the noise perceived by humans who hear the noise differs from the human auditory perception. There was a case where a problem that the tendency of the indicated value of the sound level meter was different occurred. In particular, as for the impact sound and the intermittent impact sound, the dynamic characteristics of the sound level meter have a great influence, so that noise measurement and evaluation that can reproduce the human sense of hearing cannot be performed at present.

[0004]

However, the conventional non-stationary sound measurement using a sound level meter has the following problems to be solved. That is, as a basic measurement process of the sound level meter, as shown in FIG. 7, the sound pressure measured by the microphone is measured at time intervals Δ
Assuming that the sound pressure at the i-th step discretized by t is P (i), the sound pressure is divided by the reference sound pressure (P ₀ ) to be dimensionless, squared, and multiplied by the transfer characteristic of the first-order lag system. The level L _p (i) is calculated. The time constant T of the first-order lag system is 125 m in the normal Fast mode.
sec.

Conventionally, the magnitude of the noise level is often determined based on the equivalent noise level (L _eq ) for _obtaining the energy average of the sound pressure within the measurement time based on the circuit of FIG.
However, in this method, for example, the rise, magnitude and duration of the impact noise are different, and if the energy level is the same even if the sound is completely different from the auditory sense, the sound level will naturally show the same noise level. There are deficiencies in the evaluation technology.

SUMMARY OF THE INVENTION The present invention has been made in view of the above circumstances, and can evaluate the loudness (loudness) of non-stationary noise, clarify the problem of auditory noise, and reduce noise based on human hearing. It is an object of the present invention to provide a loudness evaluation device for non-stationary noises, which makes it possible to plan and implement the method.

[0007]

According to a first aspect of the present invention, there is provided a loudness evaluation apparatus for unsteady noise, comprising: a measuring means for measuring the unsteady noise; and a frequency separating means for separating the noise measured by the measuring means for each critical band. Filter, a rectifier for rectifying the noise signal extracted by the filter, an envelope calculator for obtaining an envelope of the volume rectified by the rectifier, and a level of the envelope obtained by the envelope calculator. Level calculation means,
An auditory characteristic device for obtaining the loudness for each critical band from the level converted by the level calculating device, a characteristic extracting device for extracting a change in the loudness obtained by the auditory characteristic device as a feature amount, And a loudness display means for displaying the total loudness obtained by the non-stationary sound neural network for obtaining the total loudness based on the obtained characteristic amount.

According to a second aspect of the present invention, there is provided an unsteady noise loudness evaluation device, comprising: a measuring means for measuring the unsteady noise; a filter for separating the noise measured by the measuring means by frequency separation for each critical band; A rectifier for rectifying the noise signal extracted by the filter, an envelope calculator for obtaining an envelope of the volume rectified by the rectifier, a level calculator for converting the level of the envelope obtained by the envelope calculator, An auditory characteristic device for obtaining a loudness for each critical band from the level converted by the level calculating means, and an audio characteristic device which is identified in advance so as to nonlinearly approximate between the loudness for each critical band and the total loudness for a stationary sound, Sound for the unsteady sound based on the loudness for each critical band And Runetto, characterized in that it was and a loudness display means for displaying the overall loudness obtained by the neural net for this steady sound.

According to a third aspect of the present invention, there is provided an apparatus for evaluating loudness of unsteady noise, comprising: a measuring means for measuring the unsteady noise; a filter for separating the noise measured by the measuring means by frequency separation for each critical band; A rectifier for rectifying the noise signal extracted by the filter, an envelope calculator for obtaining an envelope of the volume rectified by the rectifier, a level calculator for converting the level of the envelope obtained by the envelope calculator, An auditory characteristic device for obtaining a loudness for each critical band from the level converted by the level calculating means, and an audio characteristic device which is identified in advance so as to nonlinearly approximate between the loudness for each critical band and the total loudness for a stationary sound, Sound for the unsteady sound based on the loudness for each critical band Lunet, loudness waveform display means for displaying the waveform of the total loudness determined by the neural network for stationary sound, loudness calculation means for processing the waveform of the total loudness determined by the neural network for stationary sound according to the application, And a loudness display means for displaying the total loudness processed by the loudness calculation means.

[0010]

Embodiments of the present invention will be described below with reference to the drawings.

(First Embodiment) FIG. 1 is a block diagram showing the configuration of a non-stationary noise loudness evaluation apparatus according to a first embodiment of the present invention.

In FIG. 1, reference numeral 1 denotes a noise measuring instrument (sound meter) for measuring mechanical noise radiated from a mechanical device, and sound pressure data measured by the noise measuring instrument 1 is digitally converted by an analog / digital converter 2. It is converted into a quantity and stored in the data storage 3. By these noise measuring instrument 1, analog / digital converter 2 and data storage 3,
Measurement and data recording of the mechanical noise of the subject are performed.

The noise data stored in the data storage 3 is input to a digital filter 4.
The digital filter 4 includes filters F ₁ , F ₂ ,..., F ₂₄ that pass through the noise data, for example, noise components in 24 critical bands in the audible frequency range. The critical band is a frequency band that is closely related to an auditory phenomenon such as human frequency discrimination, and as shown in FIG.
Are representative examples.

The output of the digital filter 4 is obtained by rectifying a passing sound pressure waveform for each band by a half-wave rectifier 5,
The volumes A ₁ , A ₂ ,..., A ₂₄ are sent to the envelope calculator 6. The envelope calculator 6 determines the envelope of the waveform (envelope), and determines the corresponding volume P ₁ , P ₂ ,
.., P ₂₄ are output to the level calculator 7. The level calculator 7 converts the level of the sound volume obtained by the envelope calculator 6 and outputs the converted levels L ₁ , L ₂ ,..., L ₂₄ to the auditory characteristic device 8. The auditory characteristic unit 8 obtains loudnesses H ₁ , H ₂ ,..., H ₂₄ for each critical band using an auditory filter simulating human auditory characteristics, and outputs the loudnesses to the feature extractor 9.

The feature extractor 9 converts loudness changes, for example, fluctuation width, frequency, fluctuation amount per time, etc., which are noticed by humans by statistical analysis, as characteristic amounts S ₁ , S ₂ ,..., S ₂₄ for non-stationary sounds. Output to the neural network device 10. The non-stationary sound neural network device 10 is identified in advance so that the relationship between the feature amounts S ₁ , S ₂ ,..., S ₂₄ and the loudness of noise can be nonlinearly approximated. Then, the total loudness Z _H can be output. The output (total loudness Z _H ) of the non-stationary sound neural network device 10 is displayed on a loudness display 11.

Next, the operation of the above embodiment will be described.

In the measuring step, the sound pressure P (i) measured by the noise measuring device 1 is immediately divided by the reference sound pressure (P ₀ ) in the i-th step discretized by the time step Δt, The change is sampled (discretized) by the analog / digital converter 2 and stored in the data storage 3. The sound pressure data stored in the data storage 3 is sent to an arithmetic processing system.

In the next operation step, the stored data is subjected to band pass (band pass) of the digital filter 4.
A passing sound pressure waveform for each critical band shown in FIG. 2 is obtained by using a filter, rectified by a half-wave rectifier 5, an envelope (envelope) of the waveform is obtained by an envelope calculator 6, and a level is calculated by a level calculator 7. Convert to waveform.

In the next evaluation step, the level waveform for each critical band is input to the auditory characteristic device 8, and as shown in FIG. 3, the masking relationship established between the level waveform and the loudness waveform, ie, a . Loudness generated by the level waveform masks subsequent level waveforms, b. The signal processing that satisfies the condition that is not recognized unless the level exceeds the loudness at that time is performed.

That is, as shown in FIG. 4, in the j-th band, if the level of the i-th step is L _j (i) and the level of the i-th step is L _j (i-1), the subtractor 21 And the level increment Δ
L _j (i) is obtained, and when there is no masking (OFF), this acts on the auditory filter 25 as it is, calculates the loudness increment Δh _j (i), and is accumulated by that time and stored in the memory 27. It is added to the instantaneous loudness h _j (i). Then, the level is compared with the level L _j (i) at that time (step A). According to the comparison result, (1) h _j (i)> L _j (i): occurrence of masking (2) L _j (i)> h _j (i): The masking release condition processing is performed by the routine within the broken line in the figure. The dashed line in the routine, for example stored from the subtracter 21 to the switching element 22 and the output [Delta] L _j (i) and Kiokuko 23 of the subtractor 21 for ON / OFF input ΔL _j (i) to Kiokuko 23 The switching operation is performed by switching the switching elements 22 and 24 according to the comparison result of the above step A. Further, the band loudness H _j (i) evaluated in the band is the maximum value 2 by comparing the magnitude of the instantaneous loudness h _j (i) and the level L _j (i) (step A).
Take 8.

As shown in FIG. 5, the auditory characteristic device 8 has the following integral-type and differential-type transfer characteristics with respect to the sound pressure level, where s is Laplace transform. This is a characteristic that works only when the ascending is performed.

The _{differential: G d (s) = K} d s / {(T d s + 1) (s 2 + 2ζω 0 s + 1)} _{integral: G i (s) = K} i / (T i s + 1) where K _d and K _i are gain constants, T _d and T _i are time constants,
ω ₀ and ζ are the natural frequency and damping ratio of the second-order lag system. As an example, in the band of the center frequency of 1 kHz, it is determined as follows.

[0023] ^{_{K d = 2 × 10 4 T}} d = 0.06 (sec) ω 0 = 6.283 × 102 (rad / sec) ζ = 0.4 K i = 1 T i = 0.08 (sec) Then, they are respectively referred to as integral type and differential type loudness, denoted as _{hj (I)} (i) and _{hj (D )} (i), and the total instantaneous loudness _hj (i) is the sum of them.

The next feature extractor 9 has a close relationship with the neural network unit 10 for non-steady sounds, neural network 10 will feature amount S _1, S ₂ becomes the input, ..., S ₂₄ and the output (Overall It must be able to express the relationship of loudness Z _H ) well. Usually, before application to this embodiment,
It is necessary to build a network through basic experiments.
This feature varies depending on the type and nature of the noise.
The variation range, frequency, and time change of the loudness for each band are obtained.

Since these series of data processing are automatically performed, if noise measurement is performed, a loudness waveform close to the actual audibility of unsteady noise such as mechanical noise is calculated, and the problem of noise is clarified. You.

According to the first embodiment, the frequency separation of the unsteady sound is performed in the critical band, which is a typical auditory characteristic,
After that, since the sound pressure waveform is half-wave rectified and attention is paid to its envelope, data handling can be facilitated. Further, in the auditory characteristic device 8, the loudness is divided into a differential type and an integral type, and the sum thereof is used as an instantaneous loudness. Processing can be performed efficiently. Further, since a neural network is used for the non-stationary sound determination unit, the entire critical band (2
Four) loudness can accurately approximate the overall effect.

(Second Embodiment) Next, a second embodiment of the present invention will be described. FIG. 6 is a block diagram showing the configuration of the loudness evaluation device for unsteady noise according to the second embodiment of the present invention. In FIG. 6, reference numeral 1 denotes a noise measuring instrument (sound level meter) for measuring mechanical noise radiated from a mechanical device. The measured sound pressure data is converted into a digital quantity by an analog / digital converter 2 and a data storage device 3 is provided. Is stored.

The sound pressure data stored in the data storage 3 is input to the digital filter 4. The digital filter 4 is composed of filters F ₁ , F ₂ ,..., F ₂₄ that pass through the sound pressure data noise components in, for example, 24 critical bands within the audible frequency range.

The output of the digital filter 4 is obtained by rectifying a passing sound pressure waveform for each band by a half-wave rectifier 5,
Volumes A ₁ , A ₂ ,..., A ₂₄ are envelope calculators 6
Sent to The envelope calculator 6 determines the envelope of the waveform, and finds the corresponding volume P ₁ , P ₂ ,
, P ₂₄ are output to the level calculator 7 and the converted levels L ₁ , L ₂ ,..., L ₂₄ are output to the auditory characteristic device 8. This auditory characteristic device 8 has a loudness H ₁ ,
H _2, ..., neural network device for the steady sound in search of H ₂₄ 1
Output to 2.

The neural network device 12 for stationary sound is
Loudness H for each critical band for stationary sound₁, H_Two,
…, H_{twenty four}And total loudness Z_HNonlinear approximation between
As previously identified. So in this case
Is the loudness H which is a function of time ₁, H_Two, ..., H_{twenty four}Strange
Loudness Z_HIs also a time function
You.

The loudness waveform display 13 displays the instantaneous loudness, and the next loudness calculator 14 performs processing such as averaging the loudness waveform according to the application. The loudness waveform processed by the loudness calculator 14 is displayed on the loudness display 11.

According to the second embodiment, the auditory characteristic device 8
Since the loudness for each critical band obtained in the above is input to the neural network device 12 for stationary sound, and the total loudness obtained by the time function and the sound pressure data are compared to identify the problem of unsteady noise. The problem of unsteady noise, that is, which part of the sound is causing discomfort to humans can be quickly identified.

[0033]

As described above in detail, according to the present invention, the frequency separation of unsteady sound is performed in the critical frequency band, which is a typical auditory characteristic, and the sound pressure waveform is half-wave rectified to detect the envelope. , After level conversion, the signal is input to the auditory characteristic device. In this auditory characteristic device, the loudness is divided into a differential type and an integral type.
The sum is used as the instantaneous loudness, and at the same time, the masking signal processing is performed by comparing the sound pressure levels at that time, and the unsteady sound is judged using a neural network. Can be accurately approximated to the overall effect.

Further, according to the present invention, the loudness for each critical band obtained by the auditory characteristic device is input to the neural network device for stationary sound, and the total loudness obtained by the time function and the sound pressure data are compared to determine the unsteady state. Since the problem of noise is specified, the problem of unsteady noise can be specified quickly. Therefore, according to the present invention, it is possible to evaluate the loudness of unsteady noise, to clarify the problem of auditory noise, and to plan and implement noise countermeasures based on human hearing.

[Brief description of the drawings]

FIG. 1 is a configuration diagram of a loudness evaluation device for unsteady noise according to a first embodiment of the present invention.

FIG. 2 is an exemplary view showing a frequency definition of a representative critical band in the auditory sense applied to the embodiment.

FIG. 3 is an exemplary view showing a masking relationship established between an envelope level waveform and a loudness waveform in the embodiment;

FIG. 4 is an internal configuration diagram of an auditory characteristic device in a j-th critical band in the embodiment.

FIG. 5 is an exemplary view showing transfer characteristics of the auditory characteristic device in the embodiment.

FIG. 6 is a configuration diagram of an unsteady noise loudness evaluation device according to a second embodiment of the present invention.

FIG. 7 is a configuration diagram of a signal processing device in a conventional sound level meter.

[Explanation of symbols]

 REFERENCE SIGNS LIST 1 noise measuring device 2 analog / digital converter 3 data storage device 4 digital filter device 5 half-wave rectifier 6 envelope calculator 7 level calculator 8 auditory characteristic device 9 feature extractor 10 neural network device for unsteady sound 11 loudness display 12 Neural network device for stationary sound 13 Loudness waveform display 14 Loudness calculator

Claims (3)

[Claims] 1. A measuring means for measuring unsteady noise, a filter for separating the noise measured by the measuring means for each critical band, and a rectifying means for rectifying a noise signal extracted by the filter. An envelope calculating means for obtaining an envelope of a volume rectified by the rectifying means; a level calculating means for converting a level of the envelope obtained by the envelope calculating means; and a level for each critical band from the level converted by the level calculating means. An auditory characteristic device for obtaining a loudness; a feature extraction unit for extracting a change in loudness obtained by the auditory characteristic device as a characteristic amount; and a non-stationary sound for obtaining an overall loudness based on the characteristic amount extracted by the characteristic extraction unit. A neural network, and a loudnet for displaying the total loudness obtained by the non-stationary sound neural net. Loudness evaluation device unsteady noise, characterized by comprising a display means. 2. A measuring means for measuring unsteady noise, a filter for separating the noise measured by the measuring means by frequency for each critical band, and a rectifying means for rectifying a noise signal taken out by the filter. An envelope calculating means for obtaining an envelope of a volume rectified by the rectifying means; a level calculating means for converting a level of the envelope obtained by the envelope calculating means; and a level for each critical band from the level converted by the level calculating means. An auditory characteristic device for determining loudness, and a non-linear characteristic based on the loudness for each critical band determined in advance so as to nonlinearly approximate between the loudness for each critical band and the total loudness for the stationary sound. A neural network for a stationary sound for obtaining a total loudness for the stationary sound, and a neural network for the stationary sound. Loudness evaluation device unsteady noise, characterized by comprising a loudness display means for displaying meta overall loudness. 3. A measuring means for measuring unsteady noise, a filter for separating the noise measured by the measuring means for each critical band, and a rectifying means for rectifying a noise signal extracted by the filter. An envelope calculating means for obtaining an envelope of a volume rectified by the rectifying means; a level calculating means for converting a level of the envelope obtained by the envelope calculating means; and a level for each critical band from the level converted by the level calculating means. An auditory characteristic device for determining loudness, and a non-linear characteristic based on the loudness for each critical band determined in advance so as to nonlinearly approximate between the loudness for each critical band and the total loudness for the stationary sound. A neural network for a stationary sound for obtaining a total loudness for the stationary sound, and a neural network for the stationary sound. Loudness waveform display means for displaying the waveform of the total loudness obtained, loudness calculating means for processing the loudness waveform of the total loudness obtained by the stationary sound neural network according to the intended use, and total loudness processed by the loudness calculating means And a loudness display means for displaying a loudness.