JP5867209B2 - Sound removal apparatus, sound inspection apparatus, sound removal method, and sound removal program - Google Patents

Description

  The disclosed technology relates to a sound removal device, a sound inspection device, a sound removal method, and a sound removal program.

  In a quality inspection in which an acoustic device that emits sound such as sound (hereinafter referred to as sound) from a speaker is an inspection target, sound emitted from the speaker is collected by a microphone. By analyzing the measurement signal obtained by sound collection, the suitability of the output characteristics of the object to be inspected, the high-frequency sound and resonance sound (so-called chatter sound) generated by the vibration of the speaker and the housing to which the speaker is attached The presence or absence is determined.

  When sound is picked up by a microphone in an environment where noise is present in the surroundings, sound is picked up including ambient noise. Therefore, since the measurement signal obtained by collecting sound includes not only the audio signal but also the noise signal, in order to perform an appropriate quality inspection, an audio signal obtained by removing the noise signal from the measurement signal is acquired. There is a need.

  As a technique for removing a noise signal from a measurement signal, there is a spectral subtraction (SS) method. In this SS method, prior to measurement of sound emitted from an object to be inspected, a sound non-output state due to only ambient noise is measured, and the amplitude spectrum of noise based on the measurement signal of only noise not including sound obtained thereby. Is estimated. Based on the estimation result, noise is removed from the measurement signal including speech by removing the amplitude spectrum of noise from the amplitude spectrum of the measurement signal including speech.

  By the way, when the inspection place is in a factory or the like, sudden noise (unsteady noise) may occur. Especially when there is no soundproofing equipment, the inspection device is easily affected by noise. The SS method is an extremely effective method for removing stationary noise (steady noise), but the SS method is non-stationary from measurement signals including speech even if unsteady noise is acquired in advance. It is difficult to remove noise.

  It is also conceivable to remove the noise amplitude spectrum obtained from the noise measurement microphone from the measurement voice signal by using a noise measurement microphone in addition to the voice measurement microphone. This technique makes it possible to remove unsteady noise.

As for noise removal, there is a technique using a microphone array in which a plurality of microphones having a predetermined correlation are arranged. Noise removal using a microphone array is performed by specifying the direction of the noise source and the magnitude of the noise from the measurement signal including the sound measured by each of the plurality of microphone arrays, and thereby the noise component from the measurement signal including the sound. Remove. Thereby, in the removal method using a microphone array, not only stationary noise but also unsteady noise can be removed.

Japanese Patent Application Laid-Open No. 07-199990 JP 2003-167484 A JP 2005-258215 A

  However, in order to remove noise components caused by noise by using a microphone for measuring noise separately from the method for collecting sound or using a microphone array, the placement accuracy of the microphone and the frequency characteristics between the microphones Matching is required. Further, the method of using a noise measurement microphone separately from the voice pickup has a problem that complicated calculation processing is required.

  The disclosed technology is intended to accurately remove sound that does not originate from an inspection target including unsteady sound when performing a quality inspection of the inspection target that outputs sound.

  In the disclosed technique, the test waveform generation unit generates a test waveform having an audible frequency and periodically having an amplitude of zero. The measurement unit picks up the first sound output from the inspection target in accordance with the inspection waveform, and generates a measurement signal including the first sound signal in accordance with the first sound. The sound extraction unit samples the measurement signal at a zero point of a waveform corresponding to the inspection waveform included in the measurement signal, and extracts a second sound signal from the measurement signal. Furthermore, the generation unit generates the first sound signal from the measurement signal by subtracting the amplitude spectrum of the second sound signal from the amplitude spectrum of the measurement signal.

  The disclosed technique has an effect that accurate sound data corresponding to the sound output from the inspection target can be obtained.

It is a functional block diagram of the inspection apparatus which concerns on 1st Embodiment. It is a diagram which shows an example of the waveform of measurement data, and the specific waveform contained in measurement data. (A) is a diagram showing an example of a waveform of measurement data, (B) is a diagram showing an example of a specific waveform, and (C) is a diagram showing a sampled signal. It is a functional block diagram of a computer that functions as an inspection device. It is a flowchart which shows an example of an audio | voice sound collection process. (A) is a flowchart which shows an example of an extraction setting process, (B) is a flowchart which shows an example of a noise signal extraction process. (A) is a diagram showing an example of a waveform of measurement data, and (B) is a diagram showing an example of a waveform of measurement data, a specific waveform, and a noise waveform of (A). It is the diagram which expanded the principal part of FIG. 7 (B), and shows an example of the waveform of a measurement data, a specific waveform, and a noise waveform. It is a functional block diagram of the inspection apparatus which concerns on 2nd Embodiment. It is a functional block diagram of the inspection apparatus which concerns on 3rd Embodiment. It is a functional block diagram of the inspection apparatus which concerns on 4th Embodiment. It is a diagram which shows an example of the test | inspection waveform which concerns on 5th Embodiment. It is a diagram which shows an example of the power spectrum of the test | inspection waveform which concerns on 4th Embodiment. (A) is a functional block diagram which shows an example of the test | inspection process part which concerns on 4th Embodiment, (B) is a diagram which shows an example of the determination in the test | inspection process part of (A). It is a flowchart which shows an example of the test | inspection waveform production | generation which concerns on 4th Embodiment. It is a flowchart which shows an example of the audio | voice data generation process which concerns on 4th Embodiment.

  Hereinafter, an example of an embodiment of the technology disclosed will be described in detail with reference to the drawings.

  [First Embodiment]

  FIG. 1 shows an inspection apparatus 10 according to the first embodiment. The inspection apparatus 10 picks up the sound output from the speaker 14 by the inspected object 12, determines the output characteristics of the inspected object 12, the presence of high-frequency sound emitted from the inspected object 12, and resonance sound (so-called chatter sound) Data for inspection such as the above is generated. In addition, the inspection apparatus 10 performs an inspection on the inspection target 12. When picking up sound emitted from the subject 12 to be inspected, ambient sounds such as noise are also picked up.

  Ambient sounds are sounds excluding the sound output by the inspection object 12 and the sound emitted by the inspection object 12 outputting sound, and include not only noise in the inspection environment but also human conversation, music, etc. . Hereinafter, the sound output from the inspection target 12 and the sound emitted from the casing of the inspection target 12 when the inspection target 12 outputs sound in the inspection environment of the inspection target are collectively referred to as “voice”. The sound excluding “speech” is collectively referred to as “noise” and described. In the disclosed technology, the voice corresponds to the first sound, and the noise corresponds to the second sound.

  In the disclosed technique, the noise component is removed from the measurement data of the measurement signal obtained by sound collection, and the data becomes a sound signal corresponding to the sound emitted from the inspected object 12 (hereinafter referred to as “voice data”). Is generated.

  In the present embodiment, a mode in which a personal computer (PC) is applied as an example of the inspection target 12 will be described. However, the inspected object 12 in the disclosed technique may be a device including the speaker 14 that outputs sound according to the input electric signal. For example, an arbitrary acoustic device such as a speaker 14 main body, a speaker system in which a plurality of speakers 14 are used integrally, a mobile phone that outputs sound from the speaker 14, an audio device, and a personal computer including the speaker 14 can be applied. When the inspection target 12 is a speaker 14 alone or a speaker system without a sound source (audio signal output source), a sound source such as an amplifier is connected for inspection.

  The inspection apparatus 10 includes an inspection waveform generation unit 16, a measurement unit 18, a noise processing unit 20, an extraction setting unit 22, and a signal generation unit 24 as a sound removal device in the disclosed technology. The inspection waveform generator 16 outputs waveform data of an inspection waveform for the object 12 to be inspected to emit a sound having a predetermined waveform, and the measurement unit 18 performs measurement by collecting the sound output from the object 12 to be inspected. Output data. The noise processing unit 20 extracts a noise component from the measurement data, and the extraction setting unit 22 sets a sampling timing when the noise processing unit 20 extracts the noise component from the measurement data. The signal generator 24 generates sound data corresponding to the sound output from the inspection target 12 by removing noise components from the measurement data.

  In addition to the above, the inspection apparatus 10 further includes an inspection processing unit 26 that inspects the inspection target 12 based on the audio signal output from the signal generation unit 24 as a sound inspection apparatus of the disclosed technology. The inspection processing unit 26 determines the output characteristics of the inspected object 12 based on the audio data obtained by the signal generating unit 24, and inspects the presence or absence of resonance sound such as high-frequency sound and chatter sound emitted from the inspected object 12. I do.

  In the disclosed technique, measurement is performed by outputting sound having a waveform generated in advance from the speaker 14 of the inspection target 12 and collecting the output sound. The waveform set in advance may be a waveform whose signal value (amplitude, sound pressure level) periodically becomes zero, and includes a waveform whose polarity changes periodically. For example, a waveform whose polarity changes by periodically crossing zero like a sine wave or cosine wave can be applied. The inspection apparatus 10 applies a sine wave having a predetermined frequency (hereinafter referred to as an inspection waveform) as an example of a sound waveform output to the inspection target 12.

  The test waveform generator 16 includes a test waveform generator 28 that generates waveform data for outputting a test waveform having a predetermined frequency as a voice, a frequency setting unit 29 that sets a frequency generated by the test waveform generator 28, and a test waveform. The inspection data storage unit 30 stores the waveform data.

  In general, the frequency of sound that can be heard by humans is in the range of 20 Hz to 20 kHz (20,000 Hz), and hereinafter, the frequency range of 20 Hz to 20 kHz is defined as the audible range. The disclosed technique may vary the frequency of the inspection waveform, or may be a constant frequency set in advance.

  The test waveform generation unit 28 generates waveform data of a test waveform having a frequency arbitrarily set within the audible range. The frequency setting unit 29 changes the frequency of the test waveform for which the waveform data is generated by the test waveform generation unit 28 continuously or stepwise in a preset step. Hereinafter, the inspection waveform of the frequency ftest is expressed as an inspection waveform ftest.

  The inspection waveform generation unit 16 stores the waveform data of the inspection waveform ftest generated by the inspection waveform generation unit 28 in the inspection data storage unit 30. In addition, the inspection waveform generation unit 16 outputs the waveform data of the inspection waveform ftest stored in the inspection data storage unit 30 to the inspection target 12 and the extraction setting unit 22.

  In the disclosed technique, when the server 32 is installed in a factory where the inspection apparatus 10 is used, the function of the inspection data storage unit 30 and the inspection waveform ftest stored in the inspection data storage unit 30 are stored in the inspection apparatus 10 The server 32 can be provided with a function to supply to the inspection object 12.

  In this case, the waveform data of the test waveform ftest generated by the test waveform generator 28 is uploaded to the server 32 and stored in the server 32. The inspection apparatus 10 and the inspected object 12 are connected to the server 32 to download the waveform data of the inspection waveform ftest from the server 32. Further, the device under test 12 and the extraction setting unit 22 can also acquire waveform data of the test waveform ftest via various storage media.

  The object to be inspected 12 uses, for example, a storage medium such as an HDD (Hard Disc Drive) or a flash memory as the waveform data storage unit 34, and stores the waveform data of the inspection waveform ftest acquired from the server 32 in the waveform data storage unit 34. Further, the test object 12 reads the waveform data of the test waveform ftest from the waveform data storage unit 34 and drives the speaker 14 via the amplifier 36, thereby outputting sound corresponding to the test waveform ftest from the speaker 14.

  The measuring unit 18 of the inspection apparatus 10 includes a microphone 38, an amplifier 40, an AD converter (Analog to Digital Converter: ADC) 42, and a preprocessing unit 44. The microphone 38 picks up the sound output from the inspection object 12 and converts it into an electric signal. The AD converter 42 receives the electric signal output from the microphone 38 via the amplifier 40, converts the electric signal into a digital signal, and measures measurement data including sound (hereinafter referred to as measurement data Sm). ) Is output. The inspection apparatus 10 can also temporarily store the measurement data Sm in a storage medium as audio recording data.

  As a method of removing noise components from the measurement data Sm, there is a method of using a noise collecting microphone in addition to a microphone for collecting voice. As a method of removing noise components from the measurement data Sm, there is a method of collecting noise together with sound by using each microphone array using a microphone array in which a plurality of microphones are arranged at a predetermined interval. Unlike these methods, the disclosed technology uses a single microphone 38 to pick up sound including sound emitted from the subject 12 to be inspected and noise in the measurement environment.

  In the disclosed technology, noise components are removed from the measurement data Sm by performing frequency component analysis using fast Fourier transform. The preprocessing unit 44 of the measurement unit 18 performs frame division on the measurement data Sm input from the AD converter 42 in time order with a processing unit in frequency component analysis described later as one frame.

  In the disclosed technique, the measurement data Sm is sampled at each zero cross point where the signal value of the audio signal component of the test waveform ftest included in the measurement data Sm becomes zero. At the zero cross point, the amplitude of the test waveform ftest is zero, and the sound pressure of the sound output from the speaker 14 is also zero.

  The extraction setting unit 22 includes a test information storage unit 46 that stores waveform data of the test waveform ftest, a band processing unit 48, a phase adjustment unit 50, and a sampling signal output unit 52. The inspection information storage unit 46 uses a storage medium such as an HDD or a flash memory, and stores the waveform data of the inspection waveform ftest acquired from the inspection waveform generation unit 16 as inspection information.

  The band processing unit 48 includes a BPF (band-pass filter) that passes a passband frequency component set to a narrow band. The pass band of the BPF is set based on the frequency of the test waveform ftest obtained from the waveform data stored in the test information storage unit 46. The pass band of the BPF in the band processing unit 48 is, for example, a range of ± 10% of the center frequency with the frequency of the test waveform ftest as a center frequency (ftest × 0.9 to ftest × 1.1 Hz with respect to the frequency ftest). Range). In the inspection apparatus 10, the frequency of the inspection waveform ftest is variable, and the band processing unit 48 changes the passband according to the frequency of the inspection waveform ftest obtained from the inspection information storage unit 46.

  The band processing unit 48 receives the measurement data Sm from the preprocessing unit 44 and thereby removes frequency components outside the pass band included in the measurement signal. The band processing unit 48 outputs a waveform (waveform data) of a signal component that approximately matches the frequency of the test waveform ftest in the measurement data Sm. This waveform is an example of a waveform corresponding to the inspection waveform included in the measurement signal in the disclosed technique, and the inspection waveform ftest can be specified in the measurement data Sm, and is hereinafter referred to as a specific waveform.

  As shown in FIG. 2, the measurement data Sm has a waveform (waveform Sm in FIG. 2) broken by including noise having a frequency different from that of the inspection waveform ftest in the inspection waveform ftest. By performing narrow band processing on the measurement data Sm, the frequency component outside the pass band is removed. Thereby, the waveform of the specific waveform Lw becomes close to the original inspection waveform ftest, and the frequency becomes the frequency of the original inspection waveform ftest.

  The phase adjustment unit 50 shown in FIG. 1 matches the phase of the sound waveform included in the measurement data Sm output from the preprocessing unit 44 to the noise processing unit 20 with respect to the specific waveform Lw that has passed through the band processing unit 48. The phase is adjusted as follows. In this phase adjustment, the origin is made coincident with the start time of the frame or a position (time) after a predetermined time from the start time as the origin, and the phase of the specific waveform Lw that has passed through the band processing unit 48 is advanced.

  In the disclosed technique, the sampling timing is detected using the specific waveform Lw by adjusting the specific waveform Lw to the measurement data Sm by phase adjustment. In addition, the disclosed technique includes performing the following processing using the inspection waveform ftest instead of the specific waveform Lw.

In the disclosed technique, as shown in FIG. 2, zero cross points t ₀ , t ₁ , t ₂ , t ₃ ,... Of the specific waveform Lw are detected in order. In the disclosed technology, the measurement data Sm is sampled at each zero cross point, thereby obtaining signal values a ₀ , a ₁ , a ₂ , a ₃ ,... For each zero cross point of the measurement data Sm. .

  The sampling signal output unit 52 shown in FIG. 1 detects the timing at which the signal value reaches the zero point from the data of the specific waveform Lw whose phase has been adjusted, and outputs a sampling signal. The specific waveform Lw is a waveform of a signal component that changes at the same frequency as the test waveform ftest in the measurement data Sm, and the signal value indicating the amplitude is zero at the zero cross point where the polarity is switched. The sampling signal output unit 52 outputs a sampling signal to the noise processing unit 20 every time a zero cross point of the specific waveform Lw is detected.

  The noise processing unit 20 includes a noise extraction unit 54, an interpolation processing unit 56, an anti-aliasing unit 58, and a frequency component analysis unit 60. The measurement data Sm is input to the noise extraction unit 54 in units of frames from the preprocessing unit 44. In addition, the sampling signal output from the sampling signal output unit 52 is input to the noise extraction unit 54. The noise extraction unit 54 samples the measurement data Sm according to the sampling signal. At this time, the noise extraction unit 54 outputs the signal value of the measurement data Sm at the sampling period by performing resampling of the measurement data Sm every time a sampling signal is input.

  In FIG. 3A, a test waveform ftest is a sine wave having a frequency of 1 kHz and an amplitude of 1.0, and a sine wave having a frequency of 100 Hz and an amplitude of 0.1 is applied as an example of a noise waveform N. A waveform synthesized with ftest is illustrated as a measurement waveform (a waveform of measurement data Sm). As shown in the figure, the waveform of the measurement data Sm changes according to the amplitude of the 100 Hz sine wave, which is the noise waveform N, of the 1 kHz inspection waveform ftest.

  FIG. 3B shows a specific waveform Lw obtained by performing band processing on the measurement data Sm with a center frequency of 1 kHz and a passband of 0.9 kHz to 1.1 kHz (1 kHz ± 0.1 kHz). Indicates. Since the specific waveform Lw has a noise component of 100 Hz (0.1 kHz), the frequency component of the noise is removed by performing band processing, and the specific waveform Lw has a frequency waveform of the inspection waveform ftest.

The sampling signal output unit 52 detects the zero cross points t ₀ , t ₁ , t ₂ , t ₃ ... Of the specific waveform Lw in order, and outputs a sampling signal for each extracted zero cross point. The period of this sampling signal is ½ times the period of the inspection waveform ftest.

By re-sampling the measurement data Sm having the waveform shown in FIG. 3A based on this sampling signal, as shown in FIG. 3C, the signal values a ₀ , a ₁ on the waveform of the measurement data Sm, a ₂ , a ₃ ,... are obtained.

The sampling signal coincides with the zero cross point of the specific waveform Lw regarded as the inspection waveform ftest on the measurement data Sm, and at the zero cross point of the inspection waveform ftest, the output of the inspection target 12 is silent. When the output of the inspected object 12 is silent, there is no high-frequency sound or vibration sound due to the sound output corresponding to the inspection waveform ftest. Therefore, the signal values a ₀ , a ₁ , a ₂ , a ₃ ,... Of the measurement data Sm at the zero cross point of the specific waveform Lw corresponding to the test waveform ftest correspond to the noise collected by the microphone 38.

  The noise extraction unit 54 illustrated in FIG. 1 resamples the measurement data Sm with a half cycle of the inspection waveform ftest included in the measurement data Sm, and outputs a signal value due to noise to the interpolation processing unit 56. The interpolation processing unit 56 generates a noise signal N from which a noise waveform corresponding to the measurement waveform indicated by the measurement data Sm is obtained by performing an interpolation process on the signal value obtained at the sampling period. The interpolation processing unit 56 may be configured to perform interpolation by a process equivalent to envelope detection.

  The anti-aliasing unit 58 includes, for example, an LPF (Low-pass filter). The anti-aliasing unit 58 obtains noise data N from which a frequency component exceeding the set frequency is removed by passing a frequency component equal to or lower than the set frequency (cut-off frequency) of the LPF. The anti-aliasing unit 58 sets the frequency of the test waveform ftest as the cut-off frequency of the LPF, and changes the cutoff by changing the test waveform ftest.

  Since the period of the specific waveform Lw (inspection waveform ftest) is twice as long as the sampling period, it is difficult to reproduce noise if the noise data N contains a frequency component equal to or higher than the frequency of the inspection waveform ftest. The anti-aliasing unit 58 removes frequency components equal to or higher than the frequency of the test waveform ftest by setting the frequency of the test waveform ftest as the cut-off frequency of the LPF, thereby improving the reproducibility of the noise data N.

  In the disclosed technique, an amplitude component for each frequency is extracted from the measurement data Sm and the noise data N by using fast Fourier transform, which is a known technique in frequency component analysis. In the disclosed technique, the noise component is removed from the measurement data Sm by subtracting the amplitude component of the noise data N from the amplitude component of the measurement data Sm for each frequency. Note that the fast Fourier transform in the disclosed technique includes using an appropriate window function.

  The frequency component analysis unit 60 performs FFT (Fast Fourier Transform) processing on the anti-aliased noise data N. Thereby, the frequency component analysis unit 60 extracts the amplitude spectrum Nr (ω) that is amplitude information corresponding to the frequency and the phase spectrum Nθ (ω) that is phase information for the noise data N. Note that ω is an angular frequency.

  The signal generation unit 24 of the inspection apparatus 10 includes a frequency component analysis unit 62, a subtraction unit 64, a signal conversion unit 66, and an audio information storage unit 68. The frequency component analysis unit 60 performs fast Fourier transform (FFT) on the measurement data Sm input for each frame. Thereby, the frequency component analysis unit 62 extracts the amplitude spectrum Smr (ω) that is the amplitude information corresponding to the frequency and the phase spectrum Smθ (ω) that is the phase information from the measurement data Sm.

  The amplitude spectrum Smr (ω) of the measurement data Sm extracted by the frequency component analysis unit 62 is input to the subtraction unit 64 together with the amplitude spectrum N (ω) of the noise data N extracted by the frequency component analysis unit 60. The subtracting unit 64 functions as a subtractor that subtracts the amplitude spectrum Nr (ω) of the noise data N from the amplitude spectrum Smr (ω) of the measurement data Sm, and removes noise below the frequency of the test waveform ftest. ) Is output.

  The signal conversion unit 66 converts the audio data S by converting the phase spectrum Smθ (ω) extracted by the frequency component analysis unit 62 and the amplitude spectrum Sr (ω) from which the noise component has been removed by the subtraction unit 64 into audio data S. Generate. The signal conversion unit 66 performs an inverse fast Fourier transform (IFFT) on the Fourier transform performed by the frequency component analysis units 60 and 62. Since the amplitude spectrum Nr (ω) of the noise data N is removed by the subtractor 64 from the amplitude spectrum Sr (ω) subjected to the inverse Fourier transform, the noise from the measurement data Sm is obtained by the Fourier inverse transform by the signal converter 66. Audio data S from which components have been removed is obtained.

  The audio information storage unit 68 uses, for example, a storage medium such as an HDD or a flash memory, and sequentially stores the audio data S generated by the signal conversion unit 66. The inspection apparatus 10 changes the frequency of the inspection waveform ftest within a preset range within the range of 20 Hz to 20 kHz, and stores the audio data S obtained for each frequency in the audio information storage unit 68. The inspection apparatus 10 uses the sound data S of 20 Hz to 20 kHz stored in the sound information storage unit 68 to determine the output characteristics of the object to be inspected 12 and the presence / absence of high-frequency sound and chatter sound. It is used for the predetermined inspection set for.

  The inspection apparatus 10 can be realized by, for example, a computer 100 illustrated in FIG. The computer 100 includes a CPU 102, a memory 104, a nonvolatile storage unit 106, a keyboard 108, a mouse 110, a display 112, and a microphone 38, and these are connected by a bus 114. The microphone 38 is connected to the bus 114 via an amplifier 40 and an AD converter 42 as shown in FIG.

  The storage unit 106 of the computer 100 shown in FIG. 4 can be realized by a storage medium such as an HDD (Hard Disk Drive) or a flash memory. The storage unit 106 stores an inspection waveform generation program 116 for causing the computer 100 to function as the inspection waveform generation unit 16 and a measurement processing program 118 for causing the computer 100 to function as the measurement unit 18. The storage unit 106 also stores a noise processing program 120 for causing the computer 100 to function as the noise processing unit 20 and an extraction setting program 122 for causing the computer 100 to function as the extraction setting unit 22. Further, the storage unit 106 stores a signal generation program 124 for causing the computer 100 to function as the signal generation unit 24 and an inspection processing program 126 for causing the computer 100 to function as the inspection processing unit 26. CPU102 reads each program from the memory | storage part 106, expand | deploys to the memory 104, and performs the process which each program 116-126 has sequentially.

  The inspection waveform generation program 116 has a frequency setting process 128 and an inspection waveform generation process 130. The CPU 102 operates as the frequency setting unit 29 by executing the frequency setting process 128, and operates as the test waveform generating unit 28 by executing the test waveform generating process 130. Further, the CPU 102 operates as a preprocessing unit 44 that divides the measurement data Sm collected from the microphone 38 and converted into a digital signal into frames by executing a process included in the measurement processing program 118.

  The extraction setting program 122 includes a band processing process 132, a phase adjustment process 134, and a sampling signal output process 136. The CPU 102 operates as the band processing unit 48 by executing the band processing process 132, operates as the phase adjustment unit 50 by executing the phase adjustment process 134, and outputs the sampling signal by executing the sampling signal output process 136. It operates as the unit 52.

  The noise processing program 120 also includes a noise extraction process 138, an interpolation processing process 140, an anti-aliasing process 142, and a frequency component analysis process 144. The CPU 102 executes a noise extraction process 138, an interpolation process 140, an anti-aliasing process 142, and a frequency component analysis process 144. As a result, the CPU 102 operates as the noise extraction unit 54, the interpolation processing unit 56, the anti-aliasing unit 58, and the frequency component analysis unit 60.

  Further, the signal generation program 124 includes a frequency component analysis process 146, a subtraction process 148, and a signal conversion process 150. The CPU 102 operates as the frequency component analysis unit 62, the subtraction unit 64, and the signal conversion unit 66 by executing the frequency component analysis process 146, the subtraction process 148, and the signal conversion process 150.

  When the inspection apparatus 10 is realized by the computer 100, the storage unit 106 stores the inspection data storage unit 30 that stores the waveform data of the inspection waveform ftest, the inspection information storage unit 46 that stores the inspection waveform information, and the audio data S. Used as the voice information storage unit 68. In this case, the inspection data storage unit 30 can be used as the inspection information storage unit 46.

  The inspection apparatus 10 can also be realized by, for example, a semiconductor integrated circuit, more specifically, an ASIC (Application Specific Integrated Circuit).

  The operation of the first embodiment will be described below.

  The inspection apparatus 10 collects the sound output from the speaker 14 of the inspection target 12 with the microphone 38 in the production factory of the inspection target 12 and analyzes the frequency component of the collected data. The suitability of the output characteristics of the object 12 is inspected. In such a factory, for example, in addition to steady noises such as air blowing sound from air conditioners, it is a relatively short time such as air blowing sound, chime sound, tape peeling sound, and mechanical movement sound. There is unsteady noise generated in the regulations. For this reason, if there is no soundproofing equipment, not only steady noise but also non-steady noise is generated, and sound is collected together with the sound output from the inspection object 12.

  The inspection apparatus 10 generates sound data S corresponding to the sound output from the inspection target 12 from the measurement data Sm obtained by collecting sound with a single microphone 38 in an environment without soundproofing equipment.

  Hereinafter, with reference to FIG. 5 to FIG. 8, generation of voice data in which noise is removed from measurement data in the inspection apparatus 10 will be described.

  When the inspection object 12 is inspected using the inspection apparatus 10, the inspection waveform generation unit 28 generates waveform data of the inspection waveform ftest of each frequency in the inspection range in advance and stores it in the inspection data storage unit 30. The frequency setting unit 29 sets the frequency of the test waveform ftest within the audible range, and the test data storage unit 30 stores the test waveform ftest for each set frequency. In the inspection apparatus 10, the waveform data of the test waveform ftest stored in the test data storage unit 30 is collected in the waveform data storage unit 34 of the inspection target 12 and the test information storage unit 46 of the extraction setting unit 22 prior to the sound collection. Store.

  The inspection object 12 outputs sound from the speaker 14 based on the waveform data of the inspection waveform ftest. At this time, when the casing of the subject 12 to be inspected vibrates, a high-frequency sound, a chatter sound, or the like may be emitted together with the sound.

  The inspection apparatus 10 starts the sound collection process by executing, for example, the flowchart of FIG. 5 in accordance with the sound output of the inspection target 12. In this flowchart, in the first step 200, the measurement unit 18 performs measurement processing. In the measurement processing, the sound to be inspected 12 picks up sound output from the speaker 14 according to the inspection waveform ftest, and generates measurement data Sm corresponding to the picked up sound. The generated measurement data Sm is sequentially divided and stored in units of frames.

  In step 202, the extraction setting unit 22 reads out measurement data Sm for one frame, and performs noise component extraction processing from the measurement data Sm. FIG. 6A shows extraction setting processing executed by the extraction setting unit 22. In this flowchart, in step 220, the bandwidth processing unit 48 performs bandwidth processing on the measurement data Sm. The frequency of the inspection waveform ftest is known from the waveform data of the inspection waveform ftest. The band processing unit 18 sets a center frequency and a pass band based on the frequency of the test waveform ftest. Further, the band processing unit 48 extracts the waveform data of the specific waveform Lw corresponding to the frequency of the test waveform ftest from the measurement data Sm by passing the measurement data Sm through the pass band set to a narrow band.

  In the next step 222, the phase adjustment unit 50 performs phase adjustment processing on the extracted waveform data of the specific waveform Lw. For example, the phase adjustment unit 50 matches the phase of the specific waveform Lw with the phase of the sound (inspected waveform ftest) included in the measurement data Sm when the noise component is extracted with reference to the start time for each frame.

  Thereafter, in step 224, the sampling signal output unit 52 performs sampling signal output processing. The sampling signal output unit 52 detects a zero cross point from the waveform data of the specific waveform Lw whose phase has been adjusted, and outputs a sampling signal each time a zero cross point is detected.

  In step 204 of the flowchart of FIG. 5, the noise extraction unit 22 performs noise component extraction processing in accordance with the processing in step 202. FIG. 6B shows noise extraction processing using a sampling signal. In step 230 of this flowchart, the noise extraction unit 54 extracts the noise component from the measurement data Sm by sampling the measurement data Sm based on the sampling signal synchronized with the specific waveform Lw included in the measurement data Sm. The cycle of the specific waveform Lw is the same as the cycle of the test waveform ftest, and the measurement data Sm is resampled at each zero cross point of the specific waveform Lw, so that there is no correlation with the sound output from the test target 12. A noise component having a signal value corresponding to the sound pressure level is obtained. Further, the sampling of the measurement data Sm is repeated at a cycle that is ½ of the cycle of the test waveform ftest.

  In step 232, interpolation processing is performed on the noise component data extracted by the interpolation processing unit 56. In step 234, the anti-aliasing unit 58 performs anti-aliasing to generate noise data N from the noise component. . In the disclosed technique, the BPF passband in the band processing and the LPF set frequency in the noise extraction can be set in advance when the microphone 38 collects the voice.

  In step 236, the frequency component analysis unit 60 performs frequency component analysis on the noise data N. The frequency component analysis 60 applies FTT to the frequency component analysis for the noise data N, and obtains an amplitude spectrum Nr (ω) and a phase spectrum Nθ (ω) from the noise waveform indicated by the noise data N.

  In step 206 of the flowchart of FIG. 5, the signal generation unit 24 executes signal generation processing in accordance with the processing of step 204. In this signal generation process, first, in step 204, the frequency component analysis unit 62 performs frequency component analysis on the measurement data Sm using FFT. The frequency component analysis unit 62 performs frequency component analysis on the measurement data Sm, thereby obtaining an amplitude spectrum Smr (ω) and a phase spectrum Smθ (ω) including an audio component and a noise component. Moreover, the frequency component analysis part 62 performs the frequency component analysis with respect to the measurement data Sm in parallel with the frequency component analysis with respect to the noise data N shown in Step 236 of FIG. Further, the frequency component analysis unit 62 can also perform frequency component analysis on the measurement data Sm prior to frequency component analysis on the noise data N.

  In the next step 208, the subtraction unit 64 performs a subtraction process for removing noise components included in the measurement data Sm. The subtracting unit 64 subtracts the amplitude spectrum Nr (ω) of the noise data N from the amplitude spectrum Smr (ω) of the measurement data Sm, and generates an amplitude spectrum Sr (ω) excluding the amplitude spectrum Nr (ω). As a result, the subtracting unit 64 obtains an audio amplitude spectrum Sr (ω) corresponding to the test waveform ftest output from the test object 12.

  The amplitude spectrum Smr (ω) is obtained by synthesizing the amplitude spectrum of noise with the amplitude spectrum of speech, and the amplitude spectrum of noise includes the amplitude spectrum of stationary noise and the amplitude spectrum of non-stationary noise. From here, in the inspection apparatus 10, the amplitude spectrum Nr (ω) of the noise data N extracted from the measurement data Sm is subtracted from the amplitude spectrum Smr (ω) of the measurement data Sm, and output from the inspection object 12. A speech amplitude spectrum Sr (ω) is extracted.

  In step 210, the signal conversion unit 66 performs a signal conversion process for converting into sound data corresponding to the sound output from the inspection target 12. The signal conversion unit 66 performs inverse Fourier transform using the amplitude spectrum Sr (ω) and the phase spectrum Smθ (ω) obtained by Fourier transform on the measurement data Sm. The voice information storage unit 68 stores the voice data S generated by the inverse Fourier transform in the signal conversion unit 66.

  Thereafter, in step 212, it is confirmed whether or not processing has been performed for all frames of the test waveform ftest having the same frequency. The inspection apparatus 10 performs the generation processing of the sound data S for all the frames of the measurement data Sm with respect to the inspection waveform ftest having the same frequency, so that an affirmative determination is made in step 212 and acquisition of the audio data N with respect to the frequency of the inspection waveform ftest. Exit. Here, the generation of the sound data S using the test waveform ftest of one frequency has been described, but the test apparatus 10 executes each of the frequencies stored in the test data storage unit 30.

FIG. 7A shows an example of a waveform measured by the inspection apparatus 10. As shown in the figure, the noise signal waveform is synthesized with the sound signal waveform corresponding to the test waveform ftest, so that the amplitude of the test waveform ftest changes according to the noise signal waveform. . The inspection apparatus 10 collects sound by the microphone 38, and acquires measurement data Sm corresponding to a signal waveform obtained by synthesizing a noise signal waveform with a sound signal waveform. FIG. 7A shows the amplitude corresponding to the frequency and the sound pressure level, the test waveform ftest is a sine wave having a frequency of 1 kHz and an amplitude of 1.0, and the signal to be noise is a frequency of 100 Hz and the amplitude is 0.25. Measurement data Sm with a sine wave and a phase difference of 45 ° (π / 4) is shown.

  FIG. 7B shows an enlarged part of the time axis of FIG. 7A, and FIG. 8 shows an enlarged part of the time axis of FIG. 7B. 7B and 8, Sm is a signal waveform based on the measurement data Sm, Lw is a specific waveform corresponding to the inspection waveform ftest, and N is a noise signal waveform.

As shown in FIGS. 7B and 8, the amplitude of the measurement data Sm is a combination of the amplitude of the specific waveform Lw and the amplitude of the noise N. Further, as shown in FIG. 8, at the zero cross points t ₀ , t ₁ , t ₂ , t ₃ ,... Of the specific waveform Lw, the phase difference between the inspection waveform ftest (specific waveform Lw) and the noise N is caused. Regardless, the amplitude of the measurement data Sm matches the amplitude of the noise N. Therefore, by sampling the measurement data Sm at the zero cross points t ₀ , t ₁ , t ₂ , t ₃ ,..., Signal values a ₀ , a ₁ , a ₂ , a ₃ , corresponding to the amplitude of the noise N are obtained. ... is obtained.

  Therefore, the inspection apparatus 10 removes noise components such as steady noise and unsteady noise in an environment where not only steady noise but also unsteady noise can occur, even in a space without soundproofing equipment, Voice data S corresponding to the voice emitted from the subject 12 can be obtained. Further, when inspecting the inspection object 12, the inspection apparatus 10 changes the frequency of the inspection waveform ftest and acquires sound data for the inspection waveform ftest for each frequency. By using the sound data S for each frequency of the inspection waveform ftest obtained in this way, it is possible to inspect the object 12 to be inspected for appropriateness of output characteristics, presence of high frequency sound, chattering sound, and the like.

  In the disclosed technique, when the inspection processing unit 26 performs Fourier transform for the analysis of the sound data S, the sound information storage unit 68 uses the amplitude spectrum and phase spectrum of the sound data, which is information before Fourier inverse transform, as sound information. Including storage.

  [Second Embodiment]

  Next, a second embodiment of the disclosed technique will be described. In the second embodiment, the same parts as those in the first embodiment are denoted by the same reference numerals, and the description thereof is omitted. Hereinafter, parts different from the first embodiment will be mainly described. .

  FIG. 9 shows an inspection apparatus 10A according to the second embodiment. The second embodiment is different in that a mutual phase matching unit 70 is provided in place of the band processing unit 48 and the phase adjustment unit 50 of the first embodiment.

  In the disclosed technique, a cross-correlation coefficient known in the art is used. In the disclosed technique, the test waveform is used to prioritize and collect the sound emitted from the inspected object 12, so that the waveform indicated by the measurement data Sm is a waveform having the same frequency as the test waveform ftest. Includes waveform Lw. In the disclosed technique, the cross-correlation coefficient is obtained by using the measurement data Sm including the specific waveform Lw and the waveform data of the known inspection waveform ftest, so that the inspection waveform ftest is not specified without specifying the specific waveform Lw. The phase is matched with the phase of the specific waveform Lw.

  The waveform data of the inspection waveform ftest and the measurement data Sm are input to the mutual phase matching unit 70 of the extraction setting unit 22A. The signal waveform indicated by the measurement data Sm includes a specific waveform Lw that is a waveform of the frequency of the inspection waveform ftest. The cross-phase matching unit 70 uses the waveform data of the inspection waveform ftest and the measurement data Sm, and obtains a cross-correlation coefficient while shifting the time of one data with the predetermined time on the measurement data Sm as the origin. The time when the maximum value of the cross-correlation coefficient thus obtained or a preset threshold value is exceeded is the time when the specific waveform Lw included in the measurement data Sm matches the phase of the test waveform ftest.

  The mutual phase matching unit 70 outputs the phase of the inspection waveform ftest in accordance with the phase of the measurement data Sm. The sampling signal output unit 52 outputs a sampling signal for each zero cross point of the inspection waveform ftest based on the waveform data of the inspection waveform ftest whose phase is matched to the measurement data Sm.

  The inspection apparatus 10A includes the cross-phase matching unit 70, so that an accurate sampling signal in the disclosed technique can be obtained without using a BPF or the like whose pass band is changed according to the frequency of the inspection waveform ftest. In addition, since the inspection apparatus 10A uses the waveform data of the inspection waveform ftest, it can accurately detect the zero cross point and sample the measurement data Sm at an appropriate timing.

  [Third Embodiment]

  Next, a third embodiment of the disclosed technique will be described. Note that in the third embodiment, the same portions as those in the first embodiment are denoted by the same reference numerals, and the description thereof is omitted. Hereinafter, portions different from those in the first embodiment will be described.

  FIG. 10 shows an inspection apparatus 10B according to the third embodiment. The third embodiment is different from the first embodiment in that the waveform data of the inspection waveform ftest is not used when a sampling signal is output. The inspected object 12 may store waveform data of a preset frequency as waveform data of the inspection waveform ftest in the waveform data storage unit 34. The disclosed technique includes providing a device corresponding to the inspection waveform generation unit 16 separately from the inspection device 10B.

  The extraction setting unit 22B of the inspection apparatus 10B includes an inspection frequency extraction unit 72 instead of the inspection information storage unit 46. The inspection frequency extraction unit 72 detects the peak frequency from the measurement data Sm when the measurement data Sm is input.

  In the disclosed technology, the noise data N is included in the measurement data Sm. However, the purpose is to collect sound output from the inspection target 12, and the inspection target 12 is more stable than steady noise even in a noisy environment. The sound pressure level of the output sound is high. Therefore, the sound (sound) output from the inspection object 12 is a sound having a frequency of a constant sound pressure level among the sounds included in the measurement data Sm, and a sound having the highest sound pressure level (amplitude). Become.

  The inspection frequency extraction unit 72 sets the frequency having the highest amplitude or the frequency having the highest average value among the frequencies having a constant sound pressure level as the peak frequency, and the peak frequency is the frequency of the specific waveform Lw corresponding to the inspection waveform ftest. Detect as.

  The band processing unit 48 sets the center frequency and pass band of the BPF based on the frequency detected by the inspection frequency extraction unit 72. Further, the anti-aliasing unit 58 of the noise processing unit 20 sets the cutoff frequency of the LPF based on the frequency detected by the inspection frequency extraction unit 72.

  Since the inspection apparatus 10B includes the inspection frequency extraction unit 72, it is not necessary to store the waveform data of the inspection waveform ftest for extracting the specific waveform Lw. Further, the inspection apparatus 10B can accurately extract the specific waveform Lw even when the frequency of the inspection waveform ftest output from the inspection target 12 is continuously changed.

  [Fourth Embodiment]

  Next, a fourth embodiment of the disclosed technique will be described. In the fourth embodiment, the same components as those in the first embodiment are denoted by the same reference numerals, and the description thereof is omitted. Hereinafter, parts different from the first embodiment will be mainly described. .

  FIG. 11 shows an inspection apparatus 10C according to the fourth embodiment. In the inspection apparatus 10C, the inspection waveform generation unit 16A is different from the inspection waveform generation unit 16 of the first embodiment.

  In the disclosed technique, when the frequency of the test waveform ftest is low, the sampling frequency of the measurement data Sm is shortened and the frequency of the removable noise is increased by synthesizing a frequency higher than the audible range.

  The inspection waveform generation unit 16A included in the inspection apparatus 10C includes an inspection waveform generation unit 28, a frequency setting unit 29, an inspection frequency determination unit 74, a carrier frequency generation unit 76, an offset unit 78, a frequency switching unit 80, a multiplication unit 82, and always on. A signal output unit 84 is included.

  The carrier frequency generator 76 generates data having a frequency outside the audible range. When the frequency of the audible range is 20 Hz to 20 kHz, the frequency outside the audible range is a frequency of 20 kHz or more, and the carrier frequency generator 76 generates data of a frequency of 21 kHz as an example. The disclosed technique includes storing data of a preset frequency and outputting the stored data. Hereinafter, the frequency data generated by the carrier frequency generation unit 76 will be referred to as a carrier wave fc.

  The offset unit 78 outputs an offset signal that is turned off at the period of the carrier wave fc based on the waveform data of the carrier wave fc output from the carrier wave frequency generation unit 76. The test waveform generator 28 outputs the waveform data of the test waveform ftest having the frequency set by the frequency setting unit 29. The frequency setting unit 29 sets the frequency of the test waveform ftest in the range of 20 Hz to 20 kHz, so that the test waveform generation unit 28 generates the frequency in a range of 20 Hz to 20 kHz.

  The inspection frequency determination unit 74 determines whether or not the frequency of the inspection waveform ftest generated by the inspection waveform generation unit 28 exceeds a preset reference frequency fs. The inspection frequency determination unit 74 determines whether or not the frequency of the inspection waveform ftest exceeds 1 kHz by setting the reference frequency fs to 1 kHz, for example. The inspection frequency determination unit 74 operates the frequency switching unit 80 based on the determination result of the frequency of the inspection waveform ftest.

  An offset signal is input from the offset unit 78 to the frequency switching unit 80. When the frequency of the test waveform ftest is equal to or lower than the reference frequency fs (fs = 1 kHz, ftest ≦ fs), the frequency determination unit 74 switches so that the frequency switching unit 80 outputs an offset signal.

  The multiplier 82 receives the waveform data of the test waveform ftest output from the test waveform generator 28. Further, when the frequency switching unit 80 outputs an offset signal, the multiplication unit 82 receives the offset signal. The multiplier 82 functions as a multiplier for the digital signal. The offset signal is a signal for setting the signal value of the test waveform ftest to zero, and the multiplier 82 receives the offset signal, so that the offset signal has a period corresponding to the waveform data of the test waveform ftest and the frequency of the carrier wave fc. And multiply. Thereby, the multiplication unit 82 outputs the waveform data of the signal waveform in which the test waveform ftest is set to zero with the signal value of the test waveform ftest in the period of the carrier wave fc.

  When the frequency of the test waveform ftest exceeds the reference frequency fs (1 kHz) (ftest> fs), the frequency determination unit 74 switches so that the frequency switching unit 80 stops the offset signal. At this time, the frequency switching unit 80 outputs a constant signal (always-on signal) input from the always-on signal output unit 84 to the multiplication unit 82. Thereby, the multiplication unit 82 outputs the waveform data of the test waveform ftest as it is without offset. The disclosed technique includes the multiplication unit 82 passing through the waveform data of the test waveform ftest without offsetting. Hereinafter, in the fourth embodiment, a waveform corresponding to the inspection waveform ftest output from the multiplier 82 is referred to as an inspection waveform Ftest.

  In the disclosed technique, when the frequency of the test waveform ftest is lower than the reference frequency fs, the test waveform Ftest obtained by offsetting the test waveform ftest with the period of the carrier wave fc is used using the carrier wave fc having a frequency higher than the audible range. The disclosed technique includes determining whether the inspection waveform ftest is equal to or higher than the reference frequency fs or lower than the reference frequency fs.

  FIG. 12 shows an example of a waveform (inspection waveform Ftest) obtained by offsetting the inspection waveform ftest with the period of the carrier wave fc. In the figure, as an example, the frequency of the test waveform ftest is 1 kHz, and the frequency of the carrier wave fc is 21 kHz.

  The inspection waveform Ftest obtained by offsetting the inspection waveform ftest with the period of the carrier wave fc is the polarity of the inspection waveform ftest, and the inspection waveform ftest changes with the frequency of the carrier wave fc. Therefore, the test waveform Ftest has a signal value (amplitude) of zero in the cycle of the carrier wave fc whose cycle is shorter than the cycle of the test waveform ftest.

  When the measurement data Sm is sampled at the zero crossing point of the inspection waveform ftest, if the frequency of the inspection waveform ftest is low, the sampling period becomes long, and it becomes difficult to remove noise having a frequency higher than that of the inspection waveform ftest. In the disclosed technique, when the frequency of the test waveform ftest is low, the period at which the signal value of the test waveform ftest becomes zero is shortened by using the carrier wave fc having a frequency exceeding the audible range that is not recognized as sound, and the sampling period is shortened. To do.

  The disclosed technique includes using the inspection waveform ftest as it is when the frequency of the inspection waveform ftest is equal to or higher than the reference frequency fs (ftest ≧ fs). Further, the disclosed technique includes shortening the sampling interval using the carrier wave fc when the frequency of the test waveform ftest is less than the reference frequency fs (ftest <fs). Furthermore, the disclosed technique includes using the inspection waveform Ftest obtained by offsetting the inspection waveform ftest by the carrier wave fc at the frequency applied to the inspection of the inspected object 12 without providing the reference frequency fs.

  The inspection apparatus 10C illustrated in FIG. 11 downloads the waveform data of the inspection waveform Ftest of each frequency stored in the inspection data storage unit 30 and the inspection information of the waveform data storage unit 34 and the extraction setting unit 22 of the inspection target 12 by downloading or the like. Store in the storage unit 46. The test object 12 outputs sound corresponding to the waveform data of the test waveform Ftest from the speaker 14.

  When the frequency of the test waveform ftest is 1 kHz or less, a sound corresponding to the waveform data of the test waveform Ftest obtained by offsetting the test waveform ftest with a carrier wave fc of 21 kHz is output. Since the frequency of the carrier wave fc exceeds the audible range, the sound output from the speaker 14 by the inspected object 12 is the sound having the frequency of the test waveform ftest. However, in the sound corresponding to the test waveform Ftest output from the test object 12, the test waveform ftest changes with the period of the carrier wave fc. Therefore, the sound collected by the inspection apparatus 10C is the sound of the inspection waveform ftest that has changed according to the carrier wave fc.

  When the frequency of the inspection waveform ftest exceeds 1 kHz, the inspection apparatus 10C functions as the inspection apparatus 10 of the first embodiment described above, and in the following, the inspection waveform Ftest with the frequency of the inspection waveform ftest of less than 1 kHz is applied. Explain the case.

  FIG. 13 shows the power spectrum of the test waveform Ftest when the frequency of the test waveform ftest is 1 kHz and the frequency of the carrier wave fc is 21 kHz. In the test waveform Ftest obtained by offsetting the test waveform ftest with the period of the carrier wave fc, the frequency (fc + ftest) obtained by adding the frequency (fc) of the carrier wave fc to the frequency (ftest) of the test waveform ftest, and the subtracted frequency (fc−ftest). The spectrum appears at.

  In the disclosed technology, the center frequency and passband of the BPF in the band processing are set based on the inspection waveform Ftest. The center frequency is the frequency of the carrier wave fc, and the passband is in the range of “frequency of carrier wave fc−frequency of test waveform ftest (fc−ftest)” to “frequency of carrier wave fc + test waveform ftest (fc + ftest)”.

  The band processing unit 48 of the inspection apparatus 10 shown in FIG. 11 sets a pass band in the range of fc ± ftest from the waveform data of the inspection waveform Ftest. The band processing unit 48 changes the pass band according to the frequency of the test waveform ftest without changing the center frequency.

  The disclosed technique does not change the passband according to the test waveform ftest, but includes the range of the reference frequency fs (Ftest ± fs). In the disclosed technique, the measurement data Sm is sampled not at the zero cross point of the inspection waveform Ftest but at the point where the signal value becomes zero in the inspection waveform Ftest (hereinafter referred to as zero point).

  The sampling signal output unit 52 detects the zero point of the signal value from the waveform data of the inspection waveform Ftest and outputs a sampling signal. At this time, since the inspection waveform Ftest is offset by the frequency of the carrier wave fc, the sampling signal output unit 52 outputs the sampling signal at the period of the carrier wave fc.

As shown in FIG. 12, the test waveform Ftest is offset from the test waveform ftest by the period of the carrier wave fc. Therefore, the zero points t ₀ , t ₁ , t ₂ , t ₃ ,... Of the test waveform Ftest appear in the period of the carrier wave fc. The sampling period of the measurement data Sm is a constant period corresponding to the frequency of the carrier wave fc regardless of the frequency of the test waveform ftest.

  The noise processing unit 20 illustrated in FIG. 11 extracts the noise data N from the measurement data Sm by the noise extraction unit 54 performing resampling with the period of the carrier wave fc. The signal generator 24 generates the audio data S by subtracting the amplitude spectrum Nr (ω) obtained from the noise data N from the amplitude spectrum Smr (ω) of the measurement data.

  In the inspection apparatus 10C, the measurement data Sm includes the frequency component of the carrier wave fc. From here, the frequency analysis part 62 removes the frequency component exceeding an audible range from measurement data Sm before the fast Fourier transformation with respect to measurement data Sm. Further, the disclosed technique includes removing the frequency component of the carrier wave fc from the measurement data Sm at an appropriate timing such as before the inverse Fourier transform.

  In the disclosed technology, the inspection for the inspection target 12 includes an arbitrary item using the audio data of the sound output from the inspection target 12.

  FIG. 14A shows an inspection processing unit 26A that inspects the suitability of output characteristics for the inspection target 12 using the audio data S corresponding to the sound output from the inspection target 12 in the inspection apparatus 10C. ing.

  The inspection processing unit 26A includes a frequency component analysis unit 86, a reference information storage unit 88, and an inspection determination unit 90. The frequency component analysis unit 86 performs frequency component analysis on the voice data S, and the reference information storage unit 88 stores reference information that serves as a reference for inspection determination on the inspection target 12. The inspection determination unit 90 determines the suitability of the output characteristics of the inspection target 12 by comparing the analysis result of the frequency component analysis unit 86 with the reference information.

  The frequency component analysis unit 86 reads out the audio data S for each frequency from the audio information storage unit 68 and extracts feature information such as the amplitude spectrum of the audio data S for each frequency by performing fast Fourier transform (FFT).

  The reference information storage unit 88 stores the amplitude spectrum (amplitude spectrum Sr (ω)) of the audio data S with respect to the feature information of the audio data S extracted by the frequency component analysis unit 86 as reference information. The reference information of the amplitude spectrum is based on the amplitude spectrum of the voice data obtained from the same product as the object 12 to be inspected and is determined to be non-defective, and an allowable value is added to the average value of the amplitude spectrum of the threshold value of the amplitude spectrum Thus, a threshold value used as a criterion for pass / fail judgment is included.

  The inspection determination unit 90 uses the reference information stored in the reference information storage unit 88 as a threshold value, and compares it with the amplitude spectrum output by the frequency component analysis unit 86 as one of the feature information of the audio data S. The disclosed technique includes storing reference information set according to the inspection item of the inspection target 12 in the reference information storage unit 88, and the inspection determination unit 90 is characterized for each inspection item of the inspection target 12. Comparing information with reference information.

  The inspection waveform generation unit 16A and the inspection processing unit 26A can be realized by, for example, a semiconductor integrated circuit, more specifically, an ASIC or the like.

  In addition, the inspection apparatus 10C can be realized using the computer 100 including the inspection waveform generation unit 16A and the inspection processing unit 26A. In this case, although illustration is omitted, the storage unit 106 stores a test waveform generation program for causing the computer 100 to function as the test waveform generation unit 16A. The inspection waveform generation program includes a frequency setting process, an inspection waveform generation process, a frequency determination process, a carrier frequency generation process, an offset process, a normally-on signal output process, a frequency switching process, and a multiplication process. The CPU 102 operates as the frequency setting unit 29 by executing the frequency setting process, and operates as the test waveform uttering unit 28 by executing the test waveform generation process. The CPU 102 operates as the frequency determination unit 74 by executing the frequency determination process, operates as the carrier frequency generation unit 76 by executing the carrier frequency generation process, and operates as the offset unit 78 by executing the offset process. Operate. Further, the CPU 102 operates as the frequency switching unit 80 by executing the frequency switching process, operates as the multiplication unit 80 by executing the multiplication process, and executes the always-on signal output process. It operates as 84.

  The storage unit 106 stores an inspection processing program. This inspection processing program has a frequency component analysis process and an inspection determination process. The CPU 102 operates as the frequency component analysis unit 86 by executing the frequency component analysis process, and operates as the inspection determination unit 90 by executing the inspection determination process.

  When the inspection processing unit 26A is realized by the computer 100, the storage unit 106 is used as a reference information storage unit 88 that stores reference data.

  Hereinafter, the generation of the waveform data of the inspection waveform Ftest and the generation of the audio data S using the generated waveform data of the inspection waveform Ftest in the inspection apparatus 10C will be described. Note that the inspection apparatus 10C generates sound data S corresponding to the frequency in the audible range by changing the frequency of the inspection waveform ftest step by step in the audible range.

  For example, the inspection apparatus 10C generates the waveform data of the inspection waveform Ftest by executing the flowchart of FIG. In the first step 240 of this flowchart, the frequency setting unit 29 sets an initial value of the test waveform ftest. The frequency setting unit 29 sets the frequency of the test waveform ftest, which is the basis of the test waveform Ftest, as an initial value at a high frequency in the audible range. The disclosed technique includes stepwise increasing the frequency of the test waveform ftest, with a low frequency in the audible range as an initial value.

  In step 242, the test waveform generator 28 generates waveform data of the test waveform ftest having the frequency set by the frequency setting unit 29. In the next step 244, the frequency determination unit 74 determines whether or not the frequency of the test waveform ftest exceeds the reference frequency fs. If the frequency of the test waveform ftest exceeds the reference frequency fs, the frequency determination unit 74 makes an affirmative determination and proceeds to step 246. In step 246, the frequency switching unit 80 performs switching so that the always-on signal output from the always-on signal output unit 84 is input to the multiplication unit 82. Thereby, the multiplication unit 82 outputs the inspection waveform ftest as it is as the inspection waveform Ftest. The inspection data storage unit 30 stores the waveform data of the inspection waveform ftest as the waveform data of the inspection waveform Ftest.

  In the next step 248, the frequency setting unit 29 confirms whether or not to generate the waveform data of the test waveform Ftest in which the frequency of the test waveform ftest is lowered. That is, the frequency setting unit 29 confirms whether or not the generation of the waveform data of the inspection waveforms Ftest of all the frequencies necessary for the inspection of the inspected object 12 has been completed. Then, the generation of the waveform data of the inspection waveform Ftest is completed.

  On the other hand, when the frequency for generating the waveform data remains, the frequency setting unit 29 makes a negative determination and proceeds to step 250. In step 250, the frequency setting unit 29 sets the frequency of the test waveform ftest to be generated to be lowered by one level. For example, when the change width of the frequency is set to the frequency Δf, the frequency setting unit 29 sets a frequency obtained by lowering the frequency Δf from the frequency ftest as a new frequency ftest. In step 242, the test waveform generation unit 28 generates waveform data of the test waveform ftest of the frequency newly set by the frequency setting unit 29.

  Here, when the frequency of the test waveform ftest is equal to or lower than the reference frequency fs, the frequency determination unit 74 makes a negative determination in step 244. Thus, when the process proceeds to step 252, the frequency switching unit 80 causes the offset unit 78 to input the offset signal generated and output at the frequency of the carrier wave fc input from the carrier frequency generation unit 76 to the multiplication unit 80. Switch. Thereby, the multiplication unit 80 multiplies the offset signal and the waveform data of the test waveform ftest, and outputs the waveform data of the test waveform Ftest obtained by offsetting the test waveform ftest with the period of the carrier wave fc. Therefore, when the frequency of the test waveform ftest is equal to or lower than the reference frequency fc, the waveform data of the test waveform Ftest is waveform data obtained by offsetting the test waveform ftest with the period of the carrier wave fc.

  The inspection object 12 stores the waveform data of each inspection waveform Ftest in the waveform data storage unit 34. Further, the test object 12 reads the waveform data of the test waveform Ftest in ascending order or ascending order of the frequency of the test waveform ftest, and outputs sound corresponding to the read waveform data of the test waveform Ftest from the speaker 14. The inspection apparatus 10 </ b> C collects sound output from the inspection target 12.

  FIG. 16 shows the flow of processing in the inspection apparatus 10C at this time. Note that the measurement processing in the inspection apparatus 10C includes the processing shown in FIGS. 5 and 6 of the first embodiment, and supplementary description of points omitted in the first embodiment is given below. .

  In this flowchart, in the first step 260, an initial value of a frequency (frequency of the inspection waveform ftest) to be inspected by the inspection apparatus 10C is set. In the next step 262, the subject 12 to be inspected outputs sound from the speaker 14 in accordance with the inspection waveform Ftest including the inspection waveform ftest having the frequency set by the inspection apparatus 10C. The measuring unit 18 measures the sound output from the inspection target 12 by executing Step 200 while outputting the sound from the speaker 14 of the inspection target 12. Thereafter, by executing Step 202 to Step 210, the voice data S corresponding to the test waveform ftest having the set frequency is generated and stored in the voice information storage unit 76.

  When the voice data corresponding to the test waveform ftest having the set frequency is generated, in step 264, the test apparatus 10C finishes generating the voice data corresponding to the test waveforms ftest of all the frequencies necessary for the test of the test object 12. Check if you did. If there remains a frequency that requires generation of the audio data, a negative determination is made in step 264 and the process proceeds to step 266, where the frequency of the test waveform ftest for generating the audio data S is set. Thereafter, the process proceeds to step 262 repeatedly, and the audio data S corresponding to the test waveform ftest is obtained for the set range of frequencies.

  When the frequency of the inspection waveform ftest included in the inspection waveform Ftest is equal to or lower than the reference frequency fs, the inspection apparatus 10C uses the inspection waveform Ftest obtained by offsetting the inspection waveform ftest with the period of the carrier wave fc. In the inspection apparatus 10C, the extraction setting unit 22 outputs a sampling signal whose sampling period is a period corresponding to the frequency of the carrier wave fc, so that the noise extraction unit 54 can output the measurement data Sm with a period corresponding to the frequency of the carrier wave fc. Sampling is performed.

  In the inspection apparatus 10C, not only noise having a frequency equal to or lower than the inspection waveform ftest but also noise having a frequency equal to or lower than ½ of the frequency of the carrier wave fc higher than the frequency of the inspection waveform ftest with respect to a frequency lower than the reference frequency fc. Can be obtained. Therefore, the inspection apparatus 10C can widen the frequency range of noise that can be removed from the measurement data Sm by using the carrier wave fc.

The inspection apparatus 10 </ b> C uses the audio data S stored in the audio information storage unit 68 to inspect the inspection target 12 by the inspection processing unit 26 </ b> A. FIG. 14B shows an example of the amplitude spectrum obtained from the threshold value st for the amplitude spectrum stored as the reference information by the reference information storage unit 88 and the sound data of the frequency f in the inspection processing unit 26A. The frequency f is the frequency of the reference waveform ftest, and the audio data Sf is the audio data S obtained from the inspection waveform of the frequency f. In FIG. 14B, the frequency f _L is the lower limit frequency (20 Hz) of the audible range, and the frequency f _H is the upper limit frequency (20 kHz) of the audible range.

  The amplitude spectrum of the audio data Sf with respect to the frequency f (indicated by the waveform Sf in the figure) has the largest amplitude at the frequency f. Further, the amplitude spectrum of the audio data Sf becomes smaller as it goes away from the frequency f, and becomes lower than the threshold value st if the inspected object 12 can be determined to be non-defective.

  In such an amplitude spectrum of the audio data Sf, when the inspected object 12 emits a high frequency sound, a chattering sound or the like, and the audio data Sf includes a noise component, the amplitude is a threshold value at a frequency higher than the frequency f. It exceeds st (waveform w indicated by a two-dot chain line in FIG. 14B). If this cause is caused by the subject 12 to be inspected, it is judged as a noise component, or it is judged that it is caused by the subject 12 to be inspected despite being a noise component. Make a decision.

In the inspection apparatus 10C, by performing the sampling of measurement data Sm with carrier fc with a frequency outside the audible range, so it removed by extracting the noise component of a frequency higher than the frequency f of the test waveform, higher than the frequency f The accuracy of the pass / fail judgment is also increased for the frequency. Therefore, the inspection processing unit 26A is provided in the inspection device 10C, so that it is possible to suppress misjudgment that determines a non-defective product as a defective product when determining the suitability of the output characteristics of the inspection target 12 based on the audio data S. Appropriate quality determination can be made.

  In the disclosed technique, instead of the band processing unit 48 and the phase adjustment unit 50, the cross-phase matching unit 70 that performs phase adjustment by cross-correlation is applied to the inspection apparatus 10C. In addition, the disclosed technique includes applying the extraction setting unit 22 </ b> A including the inspection frequency detection unit 72 to the inspection apparatus 10 </ b> C instead of the extraction setting unit 22 including the inspection information storage unit 46.

The disclosed technique is not limited to the description in the first to fourth embodiments, and any form may be used as long as each part includes a target function. In addition, all patent applications and technical documents disclosed in the patent application described in this specification include cases where individual documents, patent applications, and technical standards are specifically and individually described to be incorporated by reference. To the same extent, it is incorporated herein by reference.

10, 10A to 10C Inspection device 12 Inspected object 14 Speaker 16, 16A Inspection waveform generation unit 18 Measurement unit 20 Noise processing unit 22, 22A Extraction setting unit 24 Signal generation unit 26, 26A Inspection processing unit 28 Inspection waveform setting unit 29 Frequency Setting unit 38 Microphone 48 Band processing unit 50, 50A Phase adjustment unit 54 Noise extraction unit 60, 62 Frequency component analysis unit 64 Subtraction unit 66 Signal generation unit 70 Mutual phase matching unit 72 Inspection frequency extraction unit 74 Inspection frequency determination unit 76 Carrier frequency Generating unit 80 Frequency switching unit

Claims (13)

A test waveform generation unit that generates a test waveform having an audible frequency and an amplitude that is periodically zero;
A measurement unit that collects the first sound output from the object to be inspected according to the inspection waveform, and generates a measurement signal including the first sound signal according to the first sound;
The sound that extracts the second sound signal from the measurement signal by sampling the measurement signal at the zero point of the waveform corresponding to the inspection waveform included in the measurement signal and interpolating the signal value obtained by the sampling An extractor;
A generation unit that generates the first sound signal from the measurement signal by subtracting the amplitude spectrum of the second sound signal from the amplitude spectrum of the measurement signal;
Sound removal device including. The inspection waveform generation unit
A waveform generator for generating a sine wave having a frequency in the audible range;
A carrier wave generator for generating a carrier wave having a frequency exceeding the frequency of the audible range;
A synthesizer that multiplies the sine wave by an offset signal obtained by offsetting the carrier wave to generate a test waveform in which the amplitude becomes zero in the period of the carrier wave;
The sound removing device according to claim 1, comprising: The inspection waveform generation unit
A setting unit for setting the frequency of the sine wave generated by the waveform generation unit;
When the sine wave generated by the waveform generation unit is output as the inspection waveform for outputting the first sound, and the frequency of the sine wave generated and set by the setting unit is lower than a reference frequency a switching unit for switching to said test waveform generated by the synthesis unit is output,
The sound removing device according to claim 2, comprising: 4. The method according to claim 1, further comprising: a band processing unit that sets a center frequency and a pass band based on the frequency of the inspection waveform and extracts a waveform having a frequency corresponding to the inspection waveform from the measurement signal. sound dividing removed by device. Sound dividing removed by apparatus of claim 4 further comprising a frequency detector for detecting the frequency of the average amplitude is the test waveform maximum frequency from the measurement signal. Based on the cross-correlation between the measurement waveform based on the measurement signal and the inspection waveform, including a phase matching unit that matches the phase of the inspection waveform to the phase of the waveform corresponding to the inspection waveform included in the measurement signal,
The sound removal device according to any one of claims 1 to 3, wherein the sound extraction unit samples the measurement signal at a zero point of the inspection waveform based on the phase collated by the phase collation unit. Said signal cancellation apparatus according to any one of claims 6 請 Motomeko 1,
Based on the sound signal of the first sound generated by the generation unit, an inspection unit that inspects the sound output by the inspection target;
Sound inspection device including A test waveform generation step for generating a test waveform having an audible frequency and whose amplitude periodically becomes zero;
A measurement step of collecting the first sound output from the object to be inspected according to the inspection waveform and generating a measurement signal including the first sound signal according to the first sound;
The sound that extracts the second sound signal from the measurement signal by sampling the measurement signal at the zero point of the waveform corresponding to the inspection waveform included in the measurement signal and interpolating the signal value obtained by the sampling An extraction step;
Generating the first sound signal from the measurement signal by subtracting the amplitude spectrum of the second sound signal from the amplitude spectrum of the measurement signal;
Sound removal method including. A waveform generating step that occurs a sine wave of a frequency of the audible range,
A carrier wave generating step for generating a carrier wave having a frequency exceeding an audible frequency;
A synthesis step of generating a test waveform in which the amplitude becomes zero in the period of the carrier wave by multiplying the sine wave and an offset signal obtained by offsetting the carrier wave;
The sound removing method according to claim 8, comprising: A setting step for setting the frequency of the sine wave generated in the waveform generation step;
When the sine wave generated in the waveform generation step is output as the inspection waveform for outputting the first sound, and the frequency of the sine wave generated in the setting step is lower than a reference frequency a switching step of switching to output said test waveform generated by the combining step,
The sound removal method according to claim 9, comprising: On the computer,
A test waveform generation step for generating a test waveform having an audible frequency and whose amplitude periodically becomes zero;
A measurement step of collecting the first sound output from the object to be inspected according to the inspection waveform and generating a measurement signal including the first sound signal according to the first sound;
The sound that extracts the second sound signal from the measurement signal by sampling the measurement signal at the zero point of the waveform corresponding to the inspection waveform included in the measurement signal and interpolating the signal value obtained by the sampling An extraction step;
Generating the first sound signal from the measurement signal by subtracting the amplitude spectrum of the second sound signal from the amplitude spectrum of the measurement signal;
A sound removal program for executing processing including A waveform generating step that occurs a sine wave of a frequency of the audible range,
A carrier wave generating step for generating a carrier wave having a frequency exceeding an audible frequency;
A synthesis step of generating a test waveform in which the amplitude becomes zero in the period of the carrier wave by multiplying the sine wave and an offset signal obtained by offsetting the carrier wave;
The sound removal program according to claim 11, comprising: A setting step for setting the frequency of the sine wave generated in the waveform generation step;
When the sine wave generated in the waveform generation step is output as the inspection waveform for outputting the first sound, and the frequency of the sine wave generated in the setting step is lower than a reference frequency a switching step of switching to output said test waveform generated by the combining step,
The sound removal program according to claim 12, including: