JPH0690634B2 - Frequency discriminator - Google Patents

Description

TECHNICAL FIELD The present invention relates to a frequency discriminator capable of outputting information regarding the frequency of a wave signal in real time, which corresponds to the human sensory system.

(Prior Art) As a device for analyzing an acoustic signal in order to realize a voice input value of a computer by a discrimination function corresponding to a sensory system, a fast Fourier transform is performed on numerical data related to the acoustic signal. A device commonly called FFT is mainly used. The FFT is a device that obtains a spectrum for each frequency of an acoustic signal and extracts and displays the characteristic of the acoustic signal by three-dimensional display on a graphic display or a printer, for example.

(Problems to be Solved by the Invention) However, the above FFT has the following problems.

For example, in the auditory system, it is possible to discriminate in real time what kind of sound an acoustic signal, that is, a sound when stimulated. In the auditory system, the sound intensity resolution is 325 steps, the frequency resolution is about 1500 steps, and the time resolution is about 25 msec. Therefore, if the FFT is required to have a discrimination function equivalent to the auditory system, about 4096 calculation points are required. When calculating about 4096 points,
There is a problem that the processing time is considerably long and the followability to the temporal fluctuation of the acoustic signal is deteriorated.

Moreover, since the FFT assumes an infinite function by iterative expansion of a finite observed waveform, false information is superimposed on the obtained spectrum. In order to suppress this, a window of time must be set, which makes the arithmetic processing complicated. Moreover, there is a problem in accuracy despite complicated arithmetic processing.

Further, the FFT needs to have an analog low-pass filter as an input means, and the measured data is the one that has passed through the filter, and cannot be said to be true input data, and there is a problem of aliasing.

Therefore, there is a problem in that it is not possible to expect high FFT calculation (measurement) accuracy due to the above-mentioned problems of superimposing false information and aliasing.

In addition, since a large number of calculation points are required, a large capacity memory is required, and there is a problem that the cost of the device increases correspondingly.

Therefore, the inventor focused on the frequency from the idea that it is the best means to perform frequency feature extraction for the discrimination of the wave signal, but the FFT measurement data and the information processed from the data are There is a lot of unnecessary data, and most of the large-capacity memory is occupied by the unnecessary data,
From this, in order to perform frequency analysis, it is necessary to select, display, and evaluate the necessary information from the vast amount of information,
I realized that real-time processing is very impossible.

Further, from the viewpoint of data processing, it is difficult to think that, for example, in the auditory system, the sine component of the input (or heard) acoustic signal and the cosine component are orthogonally expanded,
Since the fundamental method of signal processing is different from that of FFT, there is a problem that the FFT cannot be required to have a discrimination function equivalent to the auditory system.

Therefore, an object of the present invention is to provide a frequency discriminator having a discrimination function corresponding to a sensory system and capable of outputting information in real time.

(Means for Solving Problems) In order to solve the above problems, the present invention has the following configuration.

That is, input means for inputting a wave signal, a predetermined number of waveform shaping means for rectifying the wave signal input from the input means, the channel shaping means being provided in a predetermined number according to the number of channels in an arbitrary frequency band, A predetermined number corresponding to the number, a comparison unit for emitting a timing pulse corresponding to the wave signal input from the input unit, and a predetermined number according to the number of channels, the waveform shaping unit or the comparison unit. Integrating means for obtaining the data of the value corresponding to the amplitude voltage of the wave signal and the period of the wave signal by integrating the signal input from the, and a predetermined number according to the number of channels, provided by the integrating means. Holding means for holding the obtained data for each channel, and the integration pulse by the timing pulse from the comparing means. And oscillating means for issuing a signal for controlling the operation timing of the holding means, from the amplitude voltage corresponding to the time for each channel sent from the holding means, and the data obtained by the integrating means, The calculation means for obtaining the sensory quantity of the wave including the frequency data corresponding to the cycle, the storage means for storing the value obtained by the calculation means, and the calculation means stored in the storage means An output means for outputting the obtained value in real time is provided.

(Operation) The operation will be described with reference to the drawings.

An acoustic signal, which is a wave signal input from the microphone 10 that is an input unit, is sorted by the bandpass filter 20 group for each frequency band (channel), and an analog switch 26 that is a waveform shaping unit and a zero-cross comparator that is a comparison unit. Input to 24. The signal input to the analog switch 26 is half-wave rectified, and a voltage corresponding to the amplitude of the half-wave signal is output in the integrator 28, which is an integrating means, and is held by the sample & hold circuit 34, which is a holding means. On the other hand, the signals input to the zero cross comparator 24 are compared with each other, a zero crossing wave is output, a voltage corresponding to a half cycle is output in the integrator 30 which is an integrating means, and the sample & hold circuit 36 which is a holding means. Held in. The operation timings of the integrators 28 and 30 and the sample and hold circuits 34 and 36 are controlled by the pulse of the monostable multivibrator 32 which is an oscillating means by the input from the zero cross comparator 24. Sample and hold circuit 34,
The voltage data held in 36 is sent to the CPU 40 including a calculation means, processed, and the input acoustic signal is replaced with a parameter indicating the amount of sound such as the loudness and pitch of each channel and output. The sensory quantity can be displayed on the CRT 42, which is a signal, by a distribution chart for each frequency band, a color chart, or the like.

(Examples) Hereinafter, preferred examples of the present invention will be described in detail with reference to the accompanying drawings.

First, the configuration will be described.

In the present embodiment, an apparatus having a discrimination function corresponding to an acoustic system for an acoustic signal will be described.

In FIG. 1, reference numeral 10 is a microphone that is an input means and inputs an acoustic signal.

Reference numeral 12 denotes an amplifier, which amplifies the acoustic signal input from the microphone 10. The number of stages of the amplifiers 12 ... Can be set considering the amplification factor and the like.

Reference numeral 14 denotes an attenuator, which suppresses the level of the acoustic signal input from the microphone 10 or the signal amplified by the amplifier 12 ...

16 is a voltage control oscillator (VC)
O) is an oscillator for outputting an analog signal for automatically calibrating the zero adjustment and drift adjustment of the analog processing unit (integrator), and the frequency is adjusted by the output signal of the D / A converter 18 described later.

Reference numeral 20 is a band pass filter (BPF) that extracts the acoustic signal input from the microphone 10 for each arbitrary frequency band. Therefore, the number of channels for data processing is determined by the number of the bandpass filters 20. In the case of the present embodiment, in order to correspond to the auditory system, 24 channels are set to be 24 channels corresponding to the seaside bandwidth of the auditory system. 20 groups of band pass filters are provided. In FIG. 1, only one module is shown between the bandpass filter 20 and the input bus to the multiplexer 22, which will be described later, but 24 modules are provided according to the number of channels.

Reference numeral 24 is a zero-cross comparator as a comparison means, which compares the amplitudes of the input signals at a zero level, "1" when the amplitude level is zero or more, and "0" when the amplitude level is less than zero. Output a zero crossing wave.

Reference numeral 26 is an analog switch that is a waveform shaping means, and uses the zero crossing wave output from the zero cross comparator 24 as a gate control signal to extract a half cycle of the input acoustic signal and perform rectification. Half-wave rectification is performed because the characteristics of hair cells that make up the auditory system are used, and the use of half-cycles makes the circuit configuration easier than extracting one cycle.

Reference numeral 28 denotes an integrator which is an integrator, which integrates the half-wave signal sent from the analog switch 26 and outputs a voltage signal indicating the value of the amplitude voltage of the input acoustic signal.

Reference numerals 29 and 31 denote analog gates, which discharge the charge voltage of the capacitors forming the integrator 28 and the integrator 30 described later by receiving a reset pulse from the monostable multivibrator 32 described later, thereby causing the integrators 28 and 30 to operate. Reset.

30 is also an integrator which is an integrator, receives the zero crossing wave from the zero cross comparator 24, and outputs a signal indicating a half cycle of the input acoustic signal.

Reference numeral 34 is a sample-and-hold circuit which is a holding means, and holds a voltage signal indicating the amplitude obtained by the integrator 28. The data output from the sample & hold circuit 34 is output in 24 channels via the data buses A1 to A24.

36 is also a sample-and-hold circuit which is a holding means, and holds the voltage signal indicating the half cycle of the input acoustic signal obtained by the integrator 30. This sample and hold circuit
The output from 36 is output to 24 channels via the data buses T1 to T24.

32 is a monostable multivibrator which is an oscillating means,
When the zero crossing wave which is the output of the zero cross comparator 24 is input, the analog gates 29 and 31 and the sample &
For the hold circuits 34 and 36, resetting the integrators 28 and 30, etc.
It oscillates a signal that suppresses the operation timing. sample&
The signals that control the operation timings of the hold circuits 34 and 36 and the integrators 28 and 30 have a predetermined time lag to hold the necessary voltage values and then reset the integrators 28 and 30 to the initial state. I am letting you.

22 is a 48-channel multiplexer and data bus
Data of 48 channels A1 to A24 and T1 to T24 are serialized in the order of, for example, A1, T1, A2, T2, ... A24, T24, and sequentially sent to an A / D converter 38 described later.

Reference numeral 38 denotes an A / D converter, which digitizes the analog signal data sequentially sent from the 48-channel multiplexer 22 and sends the digitized analog signal data to a computer system which sequentially processes the data. In this embodiment, an A / D converter with 12 bits and a maximum conversion time of 35 μsec is used.

Reference numeral 40 denotes a CPU, which is a central processing unit of a computer system that processes data. The CPU 40 includes an arithmetic means, a storage means, and a control means, and arithmetically processes the signal data input from the A / D converter 38 according to a program stored in advance, from the amplitude information for each frequency band (for each channel), For example, the amount of sensation to the acoustic signal input to the microphone 10 such as the sound level of the sound level (sone) or the periodical information such as the mel level which is the subjective judgment level of the pitch is calculated and stored. Control the data processing.
A computer (computer system) using a microprocessor may be used for data processing.
In addition, 512 Kbytes of RAM are used as the storage means of the CPU 40, and 6010 sets of memory can be used for the set data of frequency and amplitude, and it is natural that the larger the storage capacity, the more advantageous.

42 is a CRT that is an output means, and displays only the effective line spectrum for each frequency band (channel) of the acoustic signal input to the microphone 10 on the vertical axis within the 0.5% error range that is the frequency discrimination limit of the auditory system. Then, the time is displayed on the horizontal axis, and the strength of the acoustic signal is displayed for each 10 dB at the dB level (corresponding to twice the Thorn scale) for each color.

Reference numeral 44 denotes a parallel I / O means, which has a 32-bit input / output line and outputs a channel control signal of the multiplexer 22 and a control signal of a D / A converter 18 described later.

Reference numeral 18 denotes a D / A converter, which returns the output of the parallel I / O means 44 to the voltage controlled oscillator 16 as a conversion into an analog signal in order to generate the control voltage of the voltage controlled oscillator 16.

Next, the operation of the frequency discriminator corresponding to the auditory system configured as described above will be described.

The acoustic signal input to the microphone 10 is adjusted to an appropriate level by passing through the mamp 12 ... And the attenuator 14 ...
The group of band-pass filters 20 for 24 channels is entered and divided by the band-pass filter 20 corresponding to each channel.
FIG. 2A shows an acoustic signal that has passed through one bandpass filter 20. Therefore, the amplitude and frequency modulated sine wave signals as shown in FIG. 2 (a) are extracted for each channel. The signal passed through the bandpass filter 20 enters the analog switch 26 and the zero-cross comparator 24.

The signal input to the analog switch 26 is half-wave rectified and output as a half-wave signal shown in FIG.

On the other hand, the signal of the bandpass filter 20 input to the zero cross comparator 24 is compared at a zero level and a zero crossing wave shown in FIG. 2 (c) is output. This zero crossing wave outputs a pulse (+ 5V pulse) only during the time when the signal of FIG. 2 (a) is higher than the zero level.

As described above, the analog switch 26 performs the rectification operation by using the output of the zero cross comparator 24 as the gate control signal.

Further, the output of the zero-cross comparator 24 is input to the monostable multivibrator 32, and the monostable multivibrator 32 delays the holding start pulse (second pulse) of the sample & hold circuit 34 delayed by the time t ₁ from the falling edge of the zero crossing wave. Figure (d)) and its holding start pulse
The reset pulse of the integrator 28 delayed by t ₂ (FIG. 2 (e)) is oscillated. Therefore, when the half-wave signal from the analog switch 26 is input to the integrator 28, the integrator 28 integrates the half-wave signal (determines the area of the half-wave signal) and the level rises. The zero crossing wave maximum of (FIG. 2 (c)) the sample and hold circuit 34 from the falling late by the time t ₁ of the integrator 28 (at this time, the integrator 28 half-wave signal (second The amplitude value shown in FIG. 7B is held) and the holding is started. A reset pulse is input to the analog gate 29 after a time t ₂ from the start of this holding, and the analog gate 29
Will discharge the capacitor of integrator 28 and reset the value of integrator 28 to zero. The sample-and-hold circuit 34 holds the voltage indicating the maximum value (amplitude value of the half-wave signal) of the integrator 28, which was taken in before the next holding start pulse. The output of the integrator 28 is shown in FIG. 2 (f), and the output of the sample & hold circuit 34 is shown in FIG. 2 (g). In this way, the voltage indicating the amplitude value of the acoustic signal that has passed through the bandpass filter 20 is held by the sample & hold circuit 34 for each channel. The value held in the sample & hold circuit 34 is input to the computer system via the input buses A1 to A24 for computer processing.

On the other hand, the period (half period) of the half-wave signal (FIG. 2 (b)) or zero crossing wave (FIG. 2 (c) or FIG. 3 (a))
With respect to the voltage value corresponding to the signal, the zero crossing wave is input from the zero crossing comparator 24 to the integrator 30, and the integrating operation is performed (the pulse area of the zero crossing wave is obtained). The operation is the same as when the amplitude value is obtained, and the holding start pulse (FIG. 3 (b)) of the sample & hold circuit 36 is delayed by the time t ₁ from the trailing edge of the zero crossing wave, and the monostable multivibrator is operated.
The sample-and-hold circuit 36 receives the maximum value of the integrator 30 (value indicating the zero crossing wave or the half cycle of the half wave signal) from the signal 32 and holds it. Then, a reset pulse (FIG. 3 (c)) is sent from the monostable multivibrator 32 to the analog gate 31 after a delay of time t ₂ from the hold start pulse, and the capacitor of the integrator 30 is discharged to change the value of the integrator 30. Reset to zero. The sample & hold circuit 36 holds the voltage indicating the maximum value of the integrator 30 (the value indicating the zero crossing wave or the half cycle of the half wave signal) fetched before the next holding start pulse arrives.

The output of the integrator 30 is shown in FIG. 3 (d), and the output of the sample & hold circuit 36 is shown in FIG. 3 (e). In this way, the voltage indicating the half cycle of the acoustic signal that has passed through the bandpass filter 20 is held by the sample & hold circuit 36 for each channel. The value held in the sample and hold circuit 36 is input to the computer system via the input buses T1 to T24 for computer processing.

The data held in the sample and hold circuits 34 and 36 are input to the 48CH multiplexer 22 via the input buses A1 to A24 and T1 to T24. A1, T in multiplexer 22
1, A2, T2, ... A24, T24
Read in. This is because it is possible to capture more accurate data by reading the high-frequency side that changes rapidly first.

The procedure of reading this data into the CPU 40 and the arithmetic processing will be described with reference to the flowchart of FIG. First, the operator inputs a trigger code from a keyboard (not shown) (step 100). Although the trigger code is manually input, other than the manual method, for example, a method of automatically determining a trigger when an acoustic signal having a magnitude equal to or higher than a predetermined amplitude level is input may be used. When the input of the trigger is judged (step 102), the A / D converter 38 is initialized (step 104), and the input buses A1, T1, ...
-Data for 48 channels is read in the order of A24 and T24 (step 106). The read time for these 48 channels is set to 0.1 seconds in this example. However, the actual reading time is preferably 10 msec or less for human voice and 0.2 sec or less for noise so as to correspond to the time constant of the ear. This can be arbitrarily selected by the selection of IC and the program.

In this embodiment, by repeating this data reading 125 times, a total of 12.5 seconds of acoustic data can be read.

The data corresponding to the amplitude and half cycle for each channel for 12.5 seconds read 125 times (the voltage value is selected by a predetermined coefficient in the arithmetic unit of the CPU 40, arithmetic processing is performed, and the amplitude and frequency of the acoustic signal ( (The reciprocal of the period may be calculated) (step 108) The calculated amplitude and frequency data are converted into decibels by the CPU 40 for each channel, and are divided by color as shown in the following table.

Then, the frequency is output as shown in the position corresponding to the vertical axis of the display surface of the CRT 42 (step 110).

A graph showing the output is shown in FIG.

The horizontal axis represents time, which represents 12.5 seconds.

The vertical axis represents frequency, which corresponds to the human auditory system and is between 20Hz and 20Hz.
It shows up to KHz. The plotted data is displayed in different colors according to the strength (amplitude) of the sound, as described above.
Its length indicates the measured time. The analysis of this output result will be described later.

A voltage controlled oscillator that oscillates a calibration signal for zero adjustment and drift adjustment of the multiplexer 22 and integrators 28 and 30.
16 is controlled by the parallel I / O means 44. The control signal of the voltage controlled oscillator 16 is converted into an analog signal by the D / A converter 18. Voltage controlled oscillator
When the 16 oscillates the calibration signal, the switch 46 is turned on and the acoustic signal from the microphone 10 is automatically cut.

Next, the analysis of the acoustic signal will be described together with the output screen display of the CRT 42.

Fig. 5 shows the microphones of the audio signals of "Sa, Si, Su, Se, So".
The output result when inputting from 10 is shown. Since it is an audio signal, originally it is desirable to set the data acquisition of 48 channels (48 points) to 10 msec or less, but in this experiment
It took 0.1 seconds.

Looking first at the consonant sound [s], it can be seen that a strong range (formant) of 0 dB appears near 8000 Hz. Many red signals appear near 8000 Hz on an actual CRT. Vowels [a], [i], [u], [e],
The pitch (fundamental wave), which is the characteristic of voiced sound, is 180H in [o].
It appears near z.

Looking at each vowel, [a] shows a formant near 1000 Hz. The position of this formant is slightly different depending on whether it is an adult or a child or a male or female voice. The vowel [i] has a first formant near 300 Hz and a second formant near 2000 Hz.

The vowel [u] is 1st near 500Hz, 2nd and 3rd near 1200Hz.
There is a third formant near 000Hz.

The vowel [e] has a first formant near 400 Hz and a second formant near 2000 Hz.

The vowel [o] has a formant near 500 Hz to 600 Hz.

As described above, the vowels and the consonants are considerably different, and the displayed characteristic points are the same as those analyzed by a conventional voice analysis device such as a sonagraph. Therefore, not only the frequency discriminator corresponding to the auditory system of the present embodiment can be put to practical use but also it is possible to discriminate by catching the characteristics of the sound in real time.

Therefore, it is possible to discriminate what kind of sound the input sound is, by comparing with a dictionary memory that stores the characteristics of various audio signals. Further, not only the sound but also the color and intensity (brightness) of light can be discriminated as the wave signal.

Although the various preferred embodiments of the present invention have been described above, the present invention is not limited to the above-described embodiments. For example, the method of displaying an acoustic signal is not dB and frequency, but a sone (sone). The arithmetic processing of the CPU may be changed to display in units of mel or mel, or a method of adding voltage values at extremely short intervals by a counter instead of an integrator as the integrating means in the configuration may be used. Needless to say, more modifications can be made without departing from the spirit of the invention.

(Effects of the Invention) By using the frequency discriminator according to the present invention, it is possible to capture in real time the characteristics and frequency of the wave signal.
Therefore, since the characteristic of the measured wave signal itself can be displayed, it is not necessary to perform a troublesome analysis. Further, since unnecessary information is not fetched, an unnecessary large memory is not required, and it is possible to provide a device that enables rapid processing and is inexpensive in cost. If it is a wave phenomenon, it can be applied to all wave phenomena such as sound, light, and electromagnetic waves by adjusting the resolution and frequency band. Further, since the real-time discrimination function can be applied to voice recognition technology and the like, a remarkable effect that a voice input device of a computer or a computer-equipped device can be realized.

[Brief description of drawings]

FIG. 1 is a block diagram of the frequency discriminator of the embodiment, FIG. 2 (a) is an acoustic signal passed through a bandpass filter, (b) is a half-wave signal thereof, and (c) is an output of a zero-crossing comparator. Crossover wave, (d) holding start pulse for instructing the sample & hold circuit to start holding the output of the integrator for obtaining the amplitude value, (e) reset pulse for the integrator for obtaining the amplitude value, and (f) for the integrator Output signal of
(G) is the output signal of the sample and hold circuit that holds the output of the integrator, FIG. 3 (a) is the zero crossing wave that is the output of the zero-cross comparator, and (b) is the output of the integrator that determines the half cycle. Hold start pulse for instructing start to sample & hold circuit, (c) reset pulse of integrator for obtaining half cycle, (d) output signal of integrator, (e) sample & hold output of integrator The output signal of the hold circuit, FIG. 4 is a flow chart showing the procedure from CPU data acquisition to output display, and FIG. 5 is a diagram showing an example of output when a voice signal is input. 10 …… Microphone, 20 …… Band pass filter, 24 …… Zero cross comparator, 26 …… Analog switch, 28,30
...... Integrator, 32 …… Monostable multivibrator, 34,36
…… Sample and hold circuit, 40 …… CPU.

Claims (1)

[Claims] 1. Input means for inputting a wave signal, and a predetermined number of waveform shaping means for rectifying the wave signal input from the input means, the number being provided in a predetermined number according to the number of channels in an arbitrary frequency band. Comparing means, which is provided in a predetermined number according to the number of channels, for outputting a timing pulse corresponding to the wave signal input from the input means, and a predetermined number in accordance with the number of channels, and a waveform shaping means Alternatively, integrating means for integrating the signal inputted from the comparing means to obtain data of a value corresponding to the amplitude voltage of the wave signal and the cycle of the wave signal, and a predetermined number according to the number of channels, Holding means for holding the data obtained by the integrating means for each channel, and the integration by the timing pulse from the comparing means. An oscillating means for issuing a signal for controlling the operation timing of the stage and the holding means, an amplitude voltage corresponding to the time for each channel sent from the holding means, and the data obtained by the integrating means. A calculating means for obtaining a sensory amount of a wave including frequency data corresponding to a cycle, a storing means for storing a value obtained by the calculating means, and the calculating means stored in the storing means A frequency discriminator comprising: an output unit for outputting the value obtained by the method in real time.