JP2015192364A - Sound checking device, sound checking method and sound checking program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a sound checking device capable of rightly detecting noise abnormality of an object to be checked even in the case where environmental noise is present.SOLUTION: A sound checking method includes: a recording step of collecting and recording a reference sound iteratively reproduced from a speaker of the object to be checked that moves relatively to a microphone collecting sounds, and a check sound recorded during a soundless interval provided between the reference sounds, by means of the microphone; a calculation step of calculating variance of the reference sound and calculating variance of the check sound; and a determination step of comparing the variance of the reference sound and the variance of the check sound to determine whether defective noise is being outputted from the object to be tested.

Description

  The present invention relates to a voice inspection device, a voice inspection method, and a voice inspection program.

  2. Description of the Related Art Conventionally, in a device for inspecting the sound of a speaker provided in a device, a predetermined sound pattern is repeatedly output from a right speaker and a predetermined sound is output from a left speaker to a PC in which two speakers are arranged along a uniaxial direction. The voice recording pattern is repeatedly output simultaneously with the right speaker, and the recording unit performs recording between the time when the microphone is located at the center point of the two speakers and before the center point.

  Then, the main control unit inspects the two speakers based on the sound pattern recorded at the center point of the two speakers and the sound pattern recorded before the center point (see, for example, Patent Document 1). ).

JP 2013-197780 A

  However, in the conventional speaker inspection apparatus, when environmental noise exists at the inspection location, the environmental noise is superimposed on the recorded voice pattern, and it is not possible to distinguish between the speaker failure and the environmental noise in the determination of the inspection result. In some cases, a speaker failure could not be determined correctly.

  In view of this, an object of one aspect is to provide a voice inspection apparatus that can correctly detect noise abnormality of an object to be inspected even when environmental noise exists.

  In one aspect, a voice inspection method includes a reference sound repeatedly reproduced from a speaker of an inspected object that moves relative to a microphone that collects sound, and a silent section provided between the reference sound and the reference sound. A recording step of collecting and recording the inspection sound recorded during the microphone, a calculation step of calculating a variation of the reference sound, calculating a variation of the inspection sound, a variation of the reference sound, and the And a determination step of determining whether or not defective noise is being output from the device under test by comparing the fluctuation of the inspection sound.

  According to one aspect, it is possible to provide a voice inspection device that can correctly detect noise abnormality of an object to be inspected even when environmental noise exists.

Diagram showing the configuration of the voice inspection system Block diagram showing functions of voice inspection system The figure which shows the audio | voice pattern of the reference sound output from a right speaker (a), The figure which shows the audio | voice pattern of the reference sound output from a left speaker (b) The figure which shows the change of the sound accompanying the movement of the inspection object Diagram showing computation for each possible time resolution and frequency The figure which shows the maximum value calculation of silence part voice data Diagram showing volume change curve for each frequency Diagram showing volume change curve of reference sound and bad noise The figure which shows the change of the difference of the reference sound and the bad noise Diagram showing volume change curve of reference sound and environmental noise Diagram showing change in difference between reference sound and environmental noise Flow chart showing the operation of the inspection device Flow chart showing audio data analysis processing and noise abnormality detection processing Flow chart showing abnormal sound and environmental noise separation processing Flowchart showing first noise abnormality detection processing Flow chart showing detection processing of second noise abnormality

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

  First, the configuration of the voice inspection system will be described with reference to FIGS. 1 and 2. FIG. 1 is a diagram illustrating an example of a configuration of a voice inspection system. FIG. 2 is a block diagram illustrating an example of functions of the voice inspection system.

  In FIG. 1, the voice inspection system 100 includes an inspection device 1 and a conveyor 3. A microphone 11, a barcode reader 12, and a housing inspection sensor 13 are connected to the inspection device 1. The conveyor 3 moves in the right direction in the figure, and conveys the inspection object 2 placed on the conveyor 3.

  The inspected object 2 is, for example, a personal computer including a stereo speaker, and includes a right speaker 21R and a left speaker 21L, and outputs sound. In the present embodiment, an audio inspection system is exemplified in which audio output from the right speaker 21R and the left speaker 21L is an inspection target.

  Note that “sound” in the present embodiment includes sound of a specific frequency output from a speaker, speaker noise, environmental noise, etc., and sound data is in a predetermined file format for handling sound in a computer system. Recorded data. The voice inspection is to inspect abnormalities in a sound system such as a speaker of the device under test 2 and a sound IC.

  The abnormality of the sound system to be inspected in the present embodiment is, for example, generation of abnormal sound caused by a sound IC failure or noise mixing in the sound system. An abnormality in the sound system can be determined by the presence or absence of defective noise such as “petit, petite” or “hue, hugh” that occurs during silence.

  The microphone 11 collects sound, converts the collected sound into an electric signal, and inputs the electric signal to the inspection apparatus 1. For example, the microphone 11 is fixed to the upper part of the conveyor 3 so that sound output from the right speaker 21R and the left speaker 21L of the device under test 2 passing under the microphone 11 is input. The inspected object 2 moves relative to the microphone 11 by being placed on the conveyor 3. Here, the device under test 2 is placed on the conveyor 3 so that the right speaker 21R and the left speaker 21L are aligned vertically with respect to the moving direction of the conveyor 3. As a result, the right speaker 21R and the left speaker 21L sequentially pass under the microphone 11 with a time difference corresponding to the moving speed of the conveyor 3. You may enable it to adjust the attachment position of the microphone 11 according to the shape of a product, or a speaker position, for example. The inspection target 2 may be, for example, a tablet terminal, a smartphone, a mobile phone, or the like. Further, the speaker may be one inspected object 2.

  The barcode reader 12 reads information on the barcode 24 and inputs the read information to the inspection apparatus 1. The barcode 24 is, for example, a two-dimensional barcode, and the barcode 24 represents information for identifying the object to be inspected 2 such as a product number of the object to be inspected, a serial number of the object to be inspected, a product lot, and the like. The For example, the barcode 24 is affixed or tied to a transport jig for placing the test object 2 or the test object 2 (not shown). In the present embodiment, the information of the barcode 24 is read by the barcode reader 12. However, for example, information may be read by an RFID reader using an identification tag such as RFID (Radio Frequency Identification) instead of the barcode 24.

  The housing detection sensor 13 detects the housing of the inspection object 2 placed on the conveyor 3 and inputs the detection result to the inspection apparatus 1. As the case detection sensor, for example, a proximity sensor using infrared rays or ultrasonic waves can be used. The housing detection sensor 13 directly detects the housing of the device 2 to be inspected. For example, the housing detection sensor 13 may detect a conveyance jig in which the device 2 to be inspected is stored. The detection result detected by the housing detection sensor 13 can be used, for example, at the reading timing of the barcode 24 by the barcode reader 12.

  In FIG. 2, the inspection apparatus 1 includes a recording unit 14, an audio processing unit 15, a communication control unit 16, a CPU (Central Processing Unit) 17, and a storage device 18. The device under test 2 includes a speaker control unit 22 and a CPU 23. The functions of the recording unit 14, the voice processing unit 15, or the communication control unit 16 can be implemented in the inspection apparatus 1 as hardware. Moreover, the function of the recording means 14, the audio | voice processing part 15, or the communication control part 16 can be mounted in the test | inspection apparatus 1 as a program run by CPU17.

  The recording means 14 records the voice input from the microphone 11 as voice data. The recording unit 14 records audio data by generating an audio file having a predetermined file format. An audio file may be generated for each object 2 to be inspected on the conveyor 3. Further, one audio file may be generated for a plurality of objects to be inspected.

  Recording can be started in accordance with the detection of the housing detection sensor 13. For example, a predetermined time from when the case detection sensor 13 detects the object 2 to be inspected until recording of the object 2 passes under the microphone 11 or when the case detection sensor 13 is at the object 2 to be inspected. Recording may be performed until no more are detected.

  The recording unit 14 may record, for example, information on the object 2 to be inspected read by the barcode reader 12 in the generated audio file. For example, a product number or a serial number can be used as a file name of an audio file. The audio file generated by the recording unit 14 is stored in the storage device 18.

  The voice processing unit 15 performs voice processing on the voice data of the voice file generated by the recording unit 14 as “calculation processing unit” and “extraction processing unit”, calculates the volume fluctuation of the recorded voice, Extract bad noise. Details of the audio processing and extraction processing will be described later. The voice processing unit 15 can read the voice file stored in the storage device 18 and perform voice processing, and store the processing result in the storage device 18.

  The communication processing unit 16 communicates with the barcode reader 12 and the housing detection sensor 13 and receives information input from the barcode reader 12 and a detection result of the housing detection sensor 13.

  The CPU 17 controls the operation of the inspection apparatus 1. CPU17 reads the program memorize | stored in the memory | storage device 18, when each function of the recording means 14, the audio | voice processing part 15, or the communication control part 16 is mounted as a program, and performs the read program.

  The device under test 2 includes a speaker control unit 22 and a CPU 23. The speaker control unit 22 generates a reference sound and a silent section following the reference sound, which are repeated at a predetermined period described later in the description of FIG. 3, and the generated reference sound is generated at a predetermined volume with the right speaker 21R and the left speaker. Play from 21L. The CPU 17 may control other functions of the speaker control unit 22 and the inspected object 2 (not shown).

  A bar code 24 that is read by the bar code reader 12 is attached to the object to be inspected 2 or attached thereto.

  Next, the reference sound output from the device under test 2 will be described with reference to FIG. FIG. 3A is a diagram illustrating an example of the sound pattern of the reference sound output from the right speaker. FIG. 3B is a diagram illustrating an example of the sound pattern of the reference sound output from the left speaker.

  3 (a) and 3 (b), the reference sound is a sound with a constant frequency that is repeatedly output from the right speaker 21R and the left speaker 21L for T1 seconds. Between the repeated reference sound and the reference sound, a silent section of T2 seconds illustrated as a silent part is provided. The presence / absence of noise abnormality is determined by the presence / absence of noise failure in the silent section. The reference sound and the silent section are repeated at a constant cycle (T1 + T2). The recording unit 14 described with reference to FIG. 2 records the reference sound and the sound of the silent section as sound data. The silent section is originally a silent section, but in the present embodiment, the sound input and recorded from the microphone 11 in the silent section is expressed as “sound in the silent section”.

  The reference sound is reproduced simultaneously from the left and right speakers. At this time, the frequency of the reference sound for right speaker reproduction is set to a value different from the frequency of the reference sound for left speaker reproduction. For example, the frequency of the reference sound for reproducing the right speaker can be set to 1 KHz, and the frequency of the reference sound for reproducing the left speaker can be set to 2 KHz.

  Audio data in which different frequencies are recorded simultaneously is separated for each frequency by FFT (Fast Fourier Transform) described later, and the volume, that is, the amplitude of the sound, can be measured for each frequency. By setting the frequency of the reference sound for speaker reproduction to different values on the left and right, the volume of the reference sound output from the right speaker 21R and the reference sound output from the left speaker 21L simultaneously reproduced from the recorded sound data The volume can be measured separately from the volume.

  In this embodiment, two reference sounds having different frequencies are reproduced from the two left and right speakers. For example, even when there are three or more speakers, the reference sounds having different frequencies are reproduced from the respective speakers. Thus, it is possible to separate each reference sound.

  Next, the transition of the sound collected by the microphone 11 when the inspection object 2 is placed on the conveyor 3 and moves will be described with reference to FIG. FIG. 4 is a diagram showing an example of a change in volume of the collected voice accompanying the movement of the subject to be inspected.

  In FIG. 4, t <b> 1 to t <b> 5 indicate a time series when the inspection object 2 placed on the conveyor 3 moves under the microphone 11. t1 is when the right speaker 21R is on the left side of the microphone 11 in the figure. t2 is when the microphone 11 and the right speaker 21R have the shortest distance. t3 is when the microphone 11 is positioned between the right speaker 21R and the left speaker 21L. t4 is when the microphone 11 and the left speaker 21L have the shortest distance. t5 is when the left speaker 21L is on the right side of the microphone 11.

  In the graph shown in FIG. 4, SR indicates the volume of the reference sound reproduced at 1 KHz from the right speaker 21R. NR indicates the volume of defective noise generated from the right speaker 21R. SL indicates the volume of the reference sound reproduced at 2 KHz from the left speaker 21L. NL indicates the volume of defective noise generated from the left speaker 21L.

  The volume of the sound collected by the microphone 11 is inversely proportional to the square of the distance between the microphone 11 and the speaker 21 (R, L). The maximum value is obtained when the distance between the microphone 11 and the speaker 21 (R, L) is the shortest. For example, the volume of SR and NR is the maximum value at t2, and the volume of SL and NL is the maximum value at t4. Since SR and NR are sounds generated from the same right speaker 21R, the fluctuations in the volume due to the movement of the subject 2 are recorded as the same fluctuation amounts in SR and NR where the distance between the microphone 11 and the speakers 21 (R, L) is equal. The Similarly, SL and NL are recorded as the same fluctuation amount. Therefore, by comparing the fluctuation of the volume of defective noise generated from the speaker over time with the fluctuation of the volume of reference sound over time, the speaker generating the defective noise can be identified.

  Next, a method for obtaining the maximum value for each frequency in the audio processing unit 15 described in FIG. 2 will be described with reference to FIG. FIG. 5 is a diagram illustrating an example of calculation for each time resolution and frequency resolution.

  In the graph on the left side of FIG. 5, the voice processing unit 15 divides the voice data in the silent section described in FIG. Is subjected to FFT processing. At Δt1, the environmental noise of the frequency fa and the speaker noise of the frequency fc are measured (however, the environmental noise and the speaker noise are not separated at this time). At Δt2, speaker noise at frequency fc and environmental noise at frequency fd are measured. At Δt3, no sound is measured. Furthermore, at Δt4, environmental noise at frequency fb and speaker noise at frequency fc are measured.

  In the graph on the right side of FIG. 5, the audio processing unit 15 divides the frequency of the recorded audio data by the width of the frequency resolution Δf, selects a waveform having the maximum amplitude value in each Δf, It is a graph. Since the environmental noises fa, fb, and fd were measured only once between Δt1 and Δt4, the measured waveforms are recorded as they are. On the other hand, since the speaker noise of the frequency fc has been measured three times, the waveform of Δt1 having the maximum amplitude value is selected and recorded. With the above processing, the maximum amplitude value for each Δf can be obtained.

  Next, a method for repeating the calculation of the maximum value of the sound for each frequency described in FIG. 5 by the number of times of reproduction of the reference sound will be described with reference to FIG. FIG. 6 is a diagram illustrating an example of the maximum value calculation of audio data in a silent section.

  In FIG. 6, the voice processing unit 15 repeatedly processes the voice data of the silent section extracted from the voice data from S1 to SN. For example, “N” of SN represents that the reference sound is reproduced N times while one inspection object 2 passes under the microphone 11.

  The voice processing unit 15 decomposes the cut-out silent part voice data with the time resolution Δt and the frequency resolution Δf, respectively. In the silent part voice data of S1, T2 seconds of the silent part are divided by Δt to perform FFT processing, and the maximum value for each Δf is selected and stored. For example, assuming that time is possible and the frequency resolution is 1024 bits, Δt can be (2/1024) seconds obtained by dividing 2 seconds of the silent part T2 by 1024. Δf can be set to (20000/1024) Hz when the audio output band of the speaker is up to 20 KHz. The time resolution and the frequency resolution may be set according to the processing capability of the CPU 17, for example.

  The sound processing unit 15 inputs and stores the maximum value of sound at frequencies f1 to fn. The maximum value of the audio stored here may be recorded together with the elapsed time from the start of recording, for example. The sound processing unit 15 repeats the process of storing the number of reproductions of the reference sound and the maximum value.

  Next, the result of the process described with reference to FIG. 6 will be described with reference to FIG. FIG. 7 is a diagram illustrating an example of a volume change curve for each frequency.

  The graph shown in FIG. 7 shows a case where the reference sound is recorded seven times. S1 to S7 represent volume changes of the reference sound. The number of times the reference sound is measured is desirably the number of times that an envelope representing the amount of change in the reference sound is obtained. The graphs from f1 to fn are obtained by plotting the maximum value of the measured volume for each frequency over time from S1 to S7 and expressing it as an envelope. Since f1 to fn are recorded with the same amount of fluctuation as the reference sound, the source of the recorded sound is a speaker from which the reference sound is reproduced, and is assumed to be defective noise from the device under test 2. The

  On the other hand, the sound represented by the frequency fa shows a constant value over time. Therefore, it is presumed that the sound of the frequency fa is, for example, environmental noise that does not vary with the movement of the inspection object 2.

  Next, a method for extracting defective noise from the measured voice will be described with reference to FIGS. 8, 9, and 10. FIG. FIG. 8 is a diagram illustrating an example of a volume change curve of the reference sound and the defective noise. FIG. 9 is a diagram illustrating an example of a change in the difference between the reference sound and the defective noise. FIG. 10 is a diagram illustrating an example of a volume change curve of the reference sound and the environmental noise.

  In FIG. 8, graph A is an envelope showing the amplitude of the reference sound. Graph B is an envelope of the target speech for which the amount of fluctuation in speech is compared with graph A. The voice processing unit 15 calculates the difference between the graph A and the graph B. The calculated difference (A−B) between graph A and graph B is shown in FIG.

  In FIG. 9, the difference (A−B) is within the upper limit value and the lower limit value in which the amplitude is set in advance. When the difference (A−B) is within the upper limit value and the lower limit value, the sound processing unit 15 determines that the sound of the graph B is defective noise, extracts the sound of this frequency as defective noise, and performs inspection. It is determined that the body 2 has noise abnormality.

  In FIG. 10, a graph C is an envelope of a voice to be compared with the reference sound graph A. The voice processing unit 15 calculates the difference between the graph A and the graph C. The difference (AC) between the calculated graph A and graph C is shown in FIG.

  In FIG. 11, the difference (A−C) is smaller in amplitude than the preset lower limit value. The voice processing unit 15 determines that the voice of the graph C is not bad noise when the difference (A−C) deviates from between the upper limit value and the lower limit value. In the present embodiment, noise anomaly is detected based on the difference. However, for example, defective noise may be extracted by a method of obtaining an approximate line described later with reference to the follow chart of FIG.

  By the above method, defective noise can be extracted from the recorded audio data, and noise abnormality of the device under test 2 can be detected.

  The defective noise that can be extracted in this embodiment has a constant noise frequency, a constant maximum value of the noise volume, and an envelope that compares the fluctuation amount with the reference sound envelope. It occurs at a frequency that can be obtained. For example, the audio IC defect described above can be extracted because it satisfies the above conditions.

  Further, noise that can be determined as not being bad noise is environmental noise such as a human voice, chime, air gun, or the like, in which the volume of the recorded voice does not vary with the reference sound.

  Next, the operation | movement of the process of the audio | voice test | inspection in the test | inspection apparatus 1 is demonstrated using the flowchart of FIGS. FIG. 12 is a flowchart illustrating an example of the operation of the inspection apparatus. The operation of the inspection apparatus 1 shown in the following flowchart is executed by, for example, the CPU 17 described with reference to FIG.

  In FIG. 12, the detection result of the device under test 2 is read from the housing detection sensor 13 (S11), and it is determined whether or not the sensor is ON (S12). When the sensor is turned on (YES at S12), the barcode reader 12 described with reference to FIG. (S13), the reference sound repeatedly reproduced from the speaker (21R, 21L) of the device under test 2 and the sound of the silent part are collected by the microphone and recording is started (S14).

  Next, the detection result is read from the housing detection sensor 13 (S15). When the sensor is turned off (YES in S16), the recording is terminated (S17), and the recorded data is stored in the storage device 18 (S18). . Next, the stored recording data is analyzed and noise abnormality is detected (S19), and the operation is terminated.

  Next, details of step S19 will be described with reference to FIG. FIG. 13 is a flowchart illustrating an example of an audio data analysis process and a noise abnormality detection process.

  In FIG. 13, steps S21 to S30 are analysis processing for the right speaker reference sound. Steps S31 to S39 are analysis processing for the left speaker reference sound.

  First, the voice data recorded in the storage device 18 is read (S21). The audio data may be read from an external storage medium, for example.

  Next, short-time Fourier transform (FFT) with frequency resolution Δf is performed on the audio data to be processed while shifting the window function by Δt (S22). By performing the short-time Fourier transform, the processing time for the interval Δt can be shortened.

  Next, the data after the short-time Fourier transform is searched to detect the frequency component of the right speaker reference sound (1 KHz). When the reference sound is detected, the reference sound detection flag is set to ON every Δt (S23). ). If there is no detection flag (NO in S24), the reference sound from the right speaker is not reproduced, and therefore it is determined that the right speaker sound output is abnormal (S30). When one or more reference sound detection flags are ON (YES in S24), a detection counter for counting the number of detected reference sounds is set to “N = 0” (S25).

  Next, the reference sound detection flag is searched, and it is determined whether or not the detection flag is ON at Δt continuous for the same time as the output time width “T1” of the reference sound (S26). If the detection flag is ON (YES in S26), the start time of the detected reference sound and the amplitude value of the reference sound are recorded, and the detection counter N is incremented by +1 (S27). Here, the output time width is checked to prevent erroneous counting of noise having the same frequency as the reference sound.

  Steps S26 to S27 are repeated until all the detection flags for each Δt in which the reference sound has been detected are scanned (NO in S28). When all the detection flags are scanned (YES in S28), the detection counter N is 5 or more. It is checked whether or not (S29). If the detection counter is less than 5 (NO in S29), the number of plots required to draw the envelope of the reference sound described in FIG. 7 and the like cannot be obtained, and it is determined that the right speaker audio output is abnormal (S30). When the detection counter is 5 or more (YES in S29), it is determined that the reference sound to be compared can be used in the separation process of abnormal sound and environmental noise described later.

  Next, in steps S31 to S38, the same analysis processing as that in steps S23 to S30 described in the processing for analyzing the right speaker reference sound is performed on the reference sound of the left speaker. Since the description of the contents of the processing is duplicated, it will be omitted.

  Next, an abnormal sound and environmental noise are separated (S39), and the process of step S19 is terminated.

  Next, details of the separation process between the abnormal sound and the environmental noise in step S39 will be described with reference to FIG. FIG. 14 is a flowchart illustrating an example of an abnormal sound and environmental noise separation process.

  In FIG. 14, steps S41 to S49 are processes for separating abnormal sounds generated from the right speaker 21R and environmental noises. Steps S50 to S58 are a process for separating abnormal sound generated from the left speaker 21L and environmental noise.

  First, a volume change curve of the right speaker reference sound is created. The volume change curve of the reference sound is created according to the number of detected reference sounds described with reference to FIG. The change curve of the reference sound described with reference to FIG. 7 corresponds to the case where there are seven detected reference sounds S1 to S7. The volume change curve is represented by a function f (t) with respect to time t.

  Next, the silent part, which is a silent section, is cut according to the number of reference sounds (S1 to S7) (S42). The silent part is extracted by extracting the time of the silent part time T2 from the voice data, with the time obtained by adding the reference sound reproduction time T1 to the reference sound start time (H_R_T [m]) as the start position.

  Next, NS_R [m] [n], which is the storage location of the maximum amplitude value of the silent part voice data, and the variable m are cleared (S43).

  Next, the extracted sound data of the silent part from S1 to S7 is decomposed into F_R [m] [n] [o], time resolution Δt, and frequency resolution Δf as matrix-like variables described in FIG. Store. Here, m is a variable of the number of repetitions (S1 to SN) of silent part voice data. n is a variable of the number of divisions obtained by dividing the silent part with the frequency resolution Δf. For example, when dividing the frequency of the silent part by 1024, n is a value from 1 to 1024. In addition, o is a variable of the number of divisions obtained by dividing the silent part T2 by the time resolution Δt. For example, when the silent part is divided by 1024, o has a value of 1 to 1024.

  Next, the variable n is cleared, and the maximum amplitude value is calculated for each frequency Δf for one period of the sound data of the silent part stored in F_R [m] [n] [o]. The calculation result is stored in the maximum amplitude value variable NS_R [m] [n] and the maximum amplitude value time variable NST_R [m] [n] (S44). Here, NS_R [m] [n] = F_R [m] [n] [o], NST_R [m] [n] = o, where o is the time when the amplitude is maximum.

  Next, when the frequency division number n divided by the frequency resolution Δf is not the maximum value (NO in S45), n is incremented by +1 (S46), and the maximum amplitude value and the time for all the divided frequencies are obtained. Ask.

  Next, when m, which is a variable of the number of detected reference sounds, is not the maximum value (NO in S47), m is incremented by +1 (S48), and the processing from steps S44 to S46 is repeated. When m reaches the maximum value (YES in S47), defective noise for the right speaker 21R is extracted to detect noise abnormality (S49).

  Next, in steps S50 to S58, a process for separating abnormal sound generated from the left speaker 21L and environmental noise is performed. Since the processing from step S50 to S58 is the same as the processing from step S41 to S49 in the right speaker, the description thereof is omitted.

  Next, the noise abnormality detection process by the defective noise extraction process in steps S49 and S58 will be described with reference to FIGS. FIG. 15 is a flowchart illustrating an example of a first noise abnormality detection process. Furthermore, FIG. 16 is a flowchart illustrating an example of a second noise abnormality detection process.

  In FIG. 15, as illustrated in FIGS. 8 and 10, in each period from S <b> 1 to Sn, the amplitude value f () of the reference sound at the time t at which the maximum amplitude value is generated from the maximum amplitude value of the silent part for each fn. t) is subtracted (S61). For example, “NS_R [m] [n] −f (NST_R [m] [n])”. In step S61, the value of the counter a is cleared.

  Next, it is determined whether or not the subtraction result in the variable Sx from S1 to Sn is within a certain range (S62). The fixed range is, for example, a range of predetermined upper limit value and lower limit value described in FIGS. 9 and 11. When the subtraction result is within a certain range (YES in S62), the value of the counter a is incremented by +1. The determination in step S62 is repeatedly performed until all the cycles S1 to Sn are completed (S64).

  Next, it is determined whether or not the value of the counter a is equal to or greater than a predetermined ratio with respect to the number of values N from S1 to Sn (S65). If it is equal to or greater than the predetermined ratio (YES in S65), it is determined that the sound detected in the silent part is fluctuating following the fluctuation amount of the reference sound, and is generated from the speaker of the device under test 2 Defective noise can be extracted. Through the above steps, the noise abnormality of the device under test 2 is detected (S66).

  On the other hand, when it is less than the predetermined ratio (NO in S65), it is determined that the noise of the DUT 2 is not abnormal (S67). For example, when environmental noise is recorded as audio data, it can be distinguished from noise abnormality of the DUT 2.

  The determination of noise abnormality is performed for all frequencies f1 to fn (S68), and the noise abnormality detection process is terminated.

  The process of FIG. 16 is obtained by substituting the process of steps S62 to S65 surrounded by the dotted line of FIG. 15 with the process of step S69. Therefore, the description of steps with the same reference numerals is omitted.

  In FIG. 16, it is determined whether or not the approximate curve of the subtraction result obtained in step 61 is within a predetermined range (S69). In step S65 described with reference to FIG. 15, the determination is made at a ratio where the subtraction value is within the predetermined range, whereas in step 69, an approximate curve is obtained from the obtained subtraction value, and the curve is determined in advance. It is judged whether it is in the range. Since the approximation with the reference sound can be obtained by performing the determination with the approximate curve, it is possible to determine the noise abnormality by setting the approximation.

  As mentioned above, although the Example of this invention was explained in full detail, this invention is not limited to such specific embodiment, In the range of the summary of this invention described in the claim, various deformation | transformation・ Change is possible.

Regarding the above description, the following items are further disclosed.
(Appendix 1)
A reference sound repeatedly reproduced from the speaker of the inspected object that moves relative to the microphone that collects the sound, and an inspection sound recorded during a silent interval provided between the reference sound and the reference sound; A recording process of collecting and recording with the microphone;
Calculating a variation of the reference sound and calculating a variation of the inspection sound;
A voice inspection method comprising: a step of determining whether or not defective noise is output from the DUT by comparing the fluctuation of the reference sound and the fluctuation of the inspection sound.
(Appendix 2)
The voice inspection method according to appendix 1, wherein the determination step is performed based on a difference between the variation of the reference sound and the variation of the inspection sound.
(Appendix 3)
The calculation step divides the recorded test voice in a time series to calculate a maximum value of amplitude for each frequency, and stores the calculated maximum value of the amplitude in time series to change the variation of the previous test sound. The sound inspection method according to Supplementary Note 1 or 2, wherein:
(Appendix 4)
The sound inspection method according to any one of appendices 1 to 3, wherein the calculating step calculates a variation in a reference sound having a different frequency for each speaker reproduced from a plurality of speakers.
(Appendix 5)
A reference sound repeatedly reproduced from the speaker of the inspected object that moves relative to the microphone that collects the sound, and an inspection sound recorded during a silent interval provided between the reference sound and the reference sound; Collected and recorded by the microphone,
Calculate the variation of the reference sound, calculate the variation of the inspection sound,
A voice inspection program for causing a computer to execute a process of extracting defective noise from the DUT by comparing the fluctuation of the reference sound and the fluctuation of the inspection sound.
(Appendix 6)
The voice inspection program according to appendix 5, wherein the extraction process is determined based on a difference between the fluctuation of the reference sound and the fluctuation of the inspection sound.
(Appendix 7)
The calculating process divides the recorded test voice in a time series to calculate the maximum value of amplitude for each frequency, and stores the calculated maximum value of the amplitude in time series to thereby calculate the previous test sound. The voice inspection program according to appendix 5 or 6, which calculates fluctuations.
(Appendix 8)
8. The sound inspection program according to any one of appendices 5 to 7, wherein the calculating process calculates a variation in a reference sound having a different frequency for each speaker reproduced from a plurality of speakers.
(Appendix 9)
A reference sound repeatedly reproduced from the speaker of the inspected object that moves relative to the microphone that collects the sound, and an inspection sound recorded during a silent interval provided between the reference sound and the reference sound; A recording processing unit for collecting and recording the sound with the microphone;
A calculation processing unit that calculates the variation of the reference sound and calculates the variation of the inspection sound;
An audio inspection apparatus comprising: an extraction processing unit that compares the variation of the reference sound and the variation of the inspection sound and extracts defective noise from the test object.
(Appendix 10)
The voice inspection device according to appendix 9, wherein the extraction processing unit makes a determination based on a difference between the variation of the reference sound and the variation of the inspection sound.
(Appendix 11)
The calculation processing unit divides the recorded test voice in a time series to calculate a maximum value of amplitude for each frequency, and stores the calculated maximum value of the amplitude in time series to generate a pre-test sound. The voice inspection apparatus according to appendix 9 or 10, which calculates fluctuation.
(Appendix 12)
The voice inspection device according to any one of appendices 9 to 11, wherein the calculation processing unit calculates a variation in a reference sound having a different frequency for each speaker reproduced from a plurality of speakers.

DESCRIPTION OF SYMBOLS 1 Inspection apparatus 11 Microphone 12 Bar code reader 13 Case detection sensor 14 Recording means 15 Audio | voice processing part 16 Communication control part 17 CPU
18 Storage Device 2 Inspected Object 21R Right Speaker 21L Left Speaker 22 Speaker Control Unit 23 CPU
24 Barcode

Claims (6)

A reference sound repeatedly reproduced from the speaker of the inspected object that moves relative to the microphone that collects the sound, and an inspection sound recorded during a silent interval provided between the reference sound and the reference sound; A recording process of collecting and recording with the microphone;
Calculating a variation of the reference sound and calculating a variation of the inspection sound;
A voice inspection method comprising: a step of determining whether or not defective noise is output from the DUT by comparing the fluctuation of the reference sound and the fluctuation of the inspection sound.   The voice inspection method according to claim 1, wherein the determination step is performed based on a difference between a variation in the reference sound and a variation in the inspection sound.   The calculation step divides the recorded test voice in a time series to calculate a maximum value of amplitude for each frequency, and stores the calculated maximum value of the amplitude in time series to change the variation of the previous test sound. The voice inspection method according to claim 1, wherein the method is calculated.   The voice inspection method according to claim 1, wherein the calculating step calculates a change in a reference sound having a different frequency for each speaker reproduced from a plurality of speakers. A reference sound repeatedly reproduced from the speaker of the inspected object that moves relative to the microphone that collects the sound, and an inspection sound recorded during a silent interval provided between the reference sound and the reference sound; Collected and recorded by the microphone,
Calculate the variation of the reference sound, calculate the variation of the inspection sound,
A voice inspection program for causing a computer to execute a process of extracting defective noise from the DUT by comparing the fluctuation of the reference sound and the fluctuation of the inspection sound. A reference sound repeatedly reproduced from the speaker of the inspected object that moves relative to the microphone that collects the sound, and an inspection sound recorded during a silent interval provided between the reference sound and the reference sound; A recording processing unit for collecting and recording the sound with the microphone;
A calculation processing unit that calculates the variation of the reference sound and calculates the variation of the inspection sound;
An audio inspection apparatus comprising: an extraction processing unit that compares the variation of the reference sound and the variation of the inspection sound and extracts defective noise from the test object.

Images