JP2013124039A - Sound detection device, method of detecting sound and sound detection program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To accurately detect sound of a detection object from sound data without limiting a drive recorder (DR: Drive Recorder).SOLUTION: A sound detection device 100 detects a set of partial sound data whose characteristics of vehicle shake out of sound data 110a are similar to each other as a similar sound data group 110c and calculates a background noise value to be an index of sound intensity value of the background noise based on the similar sound data group 110c. Further, a sound detection section 120e detects the sound data for the detection from the sound data 110a based on the characteristics of the sound for the detection, and corrects the detection result using the background noise value.

Description

  The present invention relates to a sound detection device and the like.

  A drive recorder (DR: Drive Recorder) is used to record a dangerous scene such as a near miss or an accident. In the following description, the drive recorder is appropriately expressed as DR. Conventional DR includes, for example, a time zone for recording video data of a vehicle-mounted camera and sensor data of various sensors using acceleration of a vehicle.

  The DR that uses the acceleration of the vehicle will be described. This DR detects the time when the acceleration value of the vehicle is equal to or greater than a threshold value, and records video data, sensor data, etc. in a time zone of about several tens of seconds before and after the detected time.

  Here, the threshold value may be set lower in consideration of an error of the acceleration sensor and data recording omission. When the threshold value is set to be low in this way, DR records data even in a scene that is not so much related to danger. For example, when the vehicle passes a steep curve or passes a rough road, the acceleration exceeds the threshold value, and the DR records each data. As a result, the data recorded in the DR includes a lot of unnecessary data, and the DR user manually removes the unnecessary data. For this reason, it is required to accurately sort unnecessary data and important data.

  In addition to the above-mentioned vehicle acceleration, there is a DR that records and uses sound data around the vehicle with a single microphone installed in the vehicle in order to sort out unnecessary data and important data. The determination of whether or not to perform recording is the same as the DR without sound data, and acceleration is used. The analysis of sound data is one of the determinations when the human actually listens to the ear and removes unnecessary data manually. It stays in use.

  Here, although it is not DR, an example of the prior art of general sound analysis will be described. In the prior art, in order to detect whether a sound to be detected, for example, an alarm sound, is present in the recorded sound data, the level of the same frequency component as the specific sound of the sound to be detected is detected. It is determined whether or not the target sound data is equal to or higher than a threshold value, and the presence or absence of an alarm sound to be detected is determined. There is also a conventional technique that determines that an alarm sound to be detected is included when the sound intensity exceeds a threshold.

JP 2006-268824 A JP 2010-237781 A

  However, the above-described conventional technique has a problem that the sound to be detected cannot be detected with high accuracy when the sound data includes background noise.

  For example, a DR microphone may be placed on a handy dashboard. When a microphone is placed on the dashboard, the microphone on the dashboard and other objects, or contact sounds between other objects, caused by the shaking of the vehicle, is captured by the microphone placed nearby, causing a large background noise. It is mixed with sound data. In the prior art, since background noise caused by such shaking is not taken into account, if background noise is included in the sound data, an error occurs in sound analysis, and the detection target sound cannot be detected with high accuracy.

  The above-described problems are not limited to DR, and can occur in various techniques for detecting a sound to be detected from sound data.

  The disclosed technology has been made in view of the above, and an object thereof is to provide a sound detection device, a sound detection method, and a sound detection program that can accurately extract a sound to be extracted.

  The disclosed sound detection apparatus includes a selection unit, a threshold value calculation unit, and a sound detection unit. The selection unit, based on the sound data acquired from the microphone mounted on the moving body and the accompanying data including the characteristics of the moving body during the acquisition of the sound data, out of the sound data, A set of partial sound data in a section having similar characteristics is selected as a similar sound data group. The threshold calculation unit calculates the threshold based on the intensities of the plurality of partial sound data included in the similar sound data group. The sound detection unit detects detection sound data indicating the detection target sound data from the sound data based on the characteristics of the detection target sound, and corrects the detection sound data based on the detection sound data and the threshold value. To do.

  According to the disclosed sound detection device, there is an effect that the sound to be detected can be detected with high accuracy.

FIG. 1 is a functional block diagram illustrating the configuration of the sound detection apparatus according to the first embodiment. FIG. 2 is a diagram for explaining the processing of the normalization unit according to the first embodiment. FIG. 3 is a diagram for explaining the process 4 for calculating the background noise value. FIG. 4 is a diagram for explaining the process 5 for calculating the background noise value. FIG. 5 is a flowchart illustrating the processing procedure of the sound detection apparatus according to the first embodiment. FIG. 6 is a functional block diagram illustrating the configuration of the sound detection apparatus according to the second embodiment. FIG. 7 is a diagram for explaining processing for generating a reliability parameter. FIG. 8 is a flowchart illustrating the processing procedure of the sound detection apparatus according to the second embodiment. FIG. 9 is a functional block diagram illustrating the configuration of the sound detection apparatus according to the third embodiment. FIG. 10 is a diagram illustrating an example of a data structure of abnormal sound occurrence probability data. FIG. 11 is a flowchart illustrating the processing procedure of the sound detection apparatus according to the third embodiment. FIG. 12 is a functional block diagram illustrating the configuration of the sound detection apparatus according to the fourth embodiment. FIG. 13 is a flowchart illustrating the processing procedure of the sound detection apparatus according to the fourth embodiment. FIG. 14 is a diagram for explaining the processing of the sound detection unit. FIG. 15 is a diagram illustrating an example of a computer that executes a sound detection program.

  Hereinafter, embodiments of a sound detection device, a sound detection method, and a sound detection program disclosed in the present application will be described in detail with reference to the drawings. Note that the present invention is not limited to the embodiments.

  A sound detection apparatus according to the first embodiment will be described. For example, it is assumed that the sound detection device is installed on a moving body such as a vehicle. FIG. 1 is a functional block diagram illustrating the configuration of the sound detection apparatus according to the first embodiment. As illustrated in FIG. 1, the sound detection apparatus 100 includes a microphone 101, an acceleration sensor 102, a speed sensor 103, and a position data acquisition unit 104. The sound detection device 100 includes an interface unit 105, an input unit 106, a display unit 107, a storage unit 110, and a control unit 120.

  The microphone 101 is a microphone that collects ambient sounds. The microphone 101 outputs sound data to the control unit 120. For example, the sound data includes the intensity of sound data for each predetermined time. The microphone 101 is disposed at an arbitrary location on the vehicle. For example, the microphone 101 may be disposed inside or outside the vehicle or inside or outside the dashboard.

  The acceleration sensor 102 is a sensor that measures the acceleration of the vehicle. The acceleration sensor 102 outputs acceleration data to the control unit 120. The acceleration data includes the magnitude of acceleration every predetermined time. The acceleration sensor 102 is disposed at an arbitrary location on the vehicle.

  The speed sensor 103 is a sensor that measures the speed of the vehicle. The speed sensor 103 outputs speed data to the control unit 120. The speed data includes the magnitude of the speed for each predetermined time. The speed sensor 103 is disposed at an arbitrary location on the vehicle.

  The position data acquisition unit 104 is a processing unit that measures the position of the vehicle. For example, the position information acquisition unit 104 measures the position of the vehicle using GPS (Global Positioning System). The position data acquisition unit 104 outputs the position data to the control unit 120. The position data includes the position of the vehicle every predetermined time.

  For example, the microphone 101, the acceleration sensor 102, the speed sensor 103, and the position data acquisition unit 104 output measured data at a frequency such as 10 Hz or 30 Hz.

  The interface unit 105 is a processing unit that executes data communication with other devices. For example, detection results of abnormal sounds, sound data when abnormal sounds occur, and the like may be output to an external device via the interface unit 105.

  The input unit 106 is an input device that inputs various data to the sound detection device 100. For example, the input unit 106 corresponds to an operation panel, a touch panel, or the like. An input device such as a keyboard may be used.

  The display unit 107 is a display device that displays various data. For example, the display device 107 corresponds to a liquid crystal panel, a touch panel, or the like.

  The storage unit 110 is a storage unit that stores sound data 110a, accompanying data 110b, and a similar sound data group 110c. The storage unit 110 corresponds to, for example, a semiconductor memory device such as a random access memory (RAM), a read only memory (ROM), or a flash memory, or a storage device such as a hard disk or an optical disk.

  The sound data 110a corresponds to sound data collected by the microphone 101.

  The accompanying data 110b includes acceleration data, speed data, and position data. The acceleration data is acceleration data measured and output by the acceleration sensor 102. The speed data is speed data measured and output by the speed sensor 103. The position data is position data measured and output by the position data acquisition unit 104.

  The acceleration data, speed data, and position data of the sound data 110a and the accompanying data 110b are associated with each other according to time.

  The similar sound data group 110c includes a plurality of sound data of sections in which the characteristics of the vehicle shake are similar in the sound data 110a. The sound data of a certain section is appropriately expressed as partial sound data. The similar sound data group 110c is generated by the selection unit 120b described later.

  The control unit 120 includes a data management unit 120a, a selection unit 120b, a normalization unit 120c, a threshold value calculation unit 120d, and a sound detection unit 120e. The control unit 120 corresponds to an integrated device such as an ASIC (Application Specific Integrated Circuit) or an FPGA (Field Programmable Gate Array). Moreover, the control part 120 respond | corresponds to electronic circuits, such as CPU and MPU (Micro Processing Unit), for example.

  The data management unit 120 a is a processing unit that acquires each data from the microphone 101, the acceleration sensor 102, the speed sensor 103, and the position data acquisition unit 104 and stores the data in the storage unit 110. The data management unit 120a causes the storage unit 110 to store the sound data acquired from the microphone 101 as the sound data 110a. The data management unit 120a causes the storage unit 110 to store acceleration data, speed data, and position data acquired from the acceleration sensor 102, the speed sensor 103, and the position data acquisition unit 104, respectively, as accompanying data 110b.

  The selection unit 120b specifies the times of sections in which the characteristics of the vehicle shake are similar based on the accompanying data 110b. The selection unit 120b extracts partial sound data corresponding to the specified time zone from the sound data 110a, and causes the storage unit 110 to store a set of each partial sound data as a similar sound data group 110c.

  Here, an example of a process in which the selection unit 120b specifies a section having similar vehicle shake characteristics based on the accompanying data 110b will be described. For example, the selection unit 120b may use acceleration data in the accompanying data 110b. The selection unit 120b determines the similarity between the waveform of a certain section of acceleration data and the waveform of another section, and sets of sections in which the similarity is equal to or greater than a predetermined value have similar vehicle shake characteristics. Identifies as an interval.

  The selection unit 120b may use speed data among the accompanying data 110b. The selection unit 120b determines the similarity between the waveform of a section with speed data and the waveform of another section, and the combination of sections in which the similarity is equal to or greater than a predetermined value has similar vehicle swing characteristics. Identifies as an interval.

  The selection unit 120b may use position data in the accompanying data 110b. The selection unit 120b determines the similarity between the curvature of a certain section and the curvature of another section based on the position data, and each set of sections in which the similarity is equal to or greater than a predetermined value is a feature of vehicle shake. May be specified as similar sections. Alternatively, the selection unit 120b obtains the road shape of the road through which the vehicle has passed using road map data (not shown) and position data, and calculates the similarity between the road shape of a certain section and the road shape of another section. You may judge.

  The selection unit 120b may specify each set of sections in which the similarity is equal to or higher than a predetermined value by using acceleration data, velocity data, and position data included in the accompanying data 110b. That is, the selection unit 120b may specify a set of sections in which the degree of similarity is equal to or higher than a predetermined value in all of the acceleration data, the speed data, and the position data as a section having similar vehicle shake characteristics.

  Here, the selection unit 120b may determine the degree of similarity using a correlation coefficient, a dynamic time warping (DTW) method, a DP (Dynamic Programming) matching method, or the like. The selection unit 120b may determine the degree of similarity using other conventional techniques.

  Acceleration data, velocity data, and position data are all measured values related to vehicle behavior and are closely related to vehicle shake. For this reason, the partial sound data in a similar section in each data is highly likely to be sound data recorded when the vehicle behaves similarly, and the background noise is likely to be similar.

  The normalization unit 120c is a processing unit that normalizes the similar sound data group 110c. An example of processing that the normalization unit 120c normalizes will be described. The normalization unit 120c selects reference partial sound data from among a plurality of partial sound data included in the similar sound data group 110c. In the following description, reference partial sound data is referred to as reference sound data. The normalization unit 120c normalizes the similar sound data group 110c by adjusting the time of the remaining partial sound data included in the similar sound data group 110c so as to have the same time as the reference sound data.

  FIG. 2 is a diagram for explaining the processing of the normalization unit according to the first embodiment. Each horizontal axis in FIG. 2 is a time axis. The partial sound data 10a in FIG. 2 is assumed to be reference sound data. The partial sound data 10b1 in FIG. 2 is partial sound data other than the reference sound data. The partial sound data 10a has sound intensity values A1 to A6. Let t be the time interval between the sound intensity values of the partial sound data 10a. The partial sound data 10b1 has sound intensity values B1 to B4. Let t be the time interval between the sound intensity values of the partial sound data 10b1.

  The normalization unit 120c is configured so that the time width from the sound intensity value A1 to the sound intensity value of the partial sound data 10a is equal to the time width from the sound intensity value B1 to the sound intensity value B4 of the partial sound data 10b1. The time of the sound intensity value B4 is updated. In addition, the normalization unit 120c updates the times B2 and B3 so that the time intervals of the sound intensity values B1 to B4 are equal. When the normalization unit 120c executes such processing, the partial sound data 10b1 becomes the partial sound data 10b2.

  After generating the partial sound data 10b2, the normalizing unit 120c generates new sound intensity values P1 to P4 based on the sound intensity values B1 to B4. By performing such processing, the partial sound data 10b2 becomes the partial sound data 10b3. The time intervals of the sound intensity values A1, A2, A3, A4, A5, and A6 are equal to the time intervals of the sound intensity values B1, P1, P2, P3, P4, and B4. For example, each time interval is t.

  Processing for calculating the sound intensity values P1 to P4 will be described. Here, a process in which the normalization unit 120c calculates the sound intensity value P1 will be described. A sound intensity value B deviated from the original time is detected before and after the sound intensity value P1. In the example shown in FIG. 2, the normalization unit 120c detects the sound intensity value B1 from the front of the sound intensity value P1, and detects the sound intensity value B2 from the rear of the sound intensity value P1. The time difference between the sound intensity value B1 and the sound intensity value P1 is t1, and the time difference between the sound intensity value B2 and the sound intensity value P1 is t2. The normalization unit 120c calculates the sound intensity value P1 based on the formula (1).

  P1 = (B1 × t1 + B2 × t2) ÷ (t1 + t2) (1)

  The normalization unit 120c calculates the values of the other sound intensity values P2 to P4 by the same method as the sound intensity value P1. Although an example of time normalization of sound intensity values is shown here, normalization may also be performed using level values of each frequency of sound. For example, the normalization unit 120c may add a normalization process for each frequency to the level value group of each frequency recorded at regular sampling time intervals.

  The threshold calculation unit 120d is a processing unit that calculates a threshold for determining background noise based on the normalized similar sound data group 110c. Hereinafter, a threshold for determining background noise is referred to as a background noise value. The threshold value calculation unit 120d outputs the background noise value to the sound detection unit 120e.

  Here, the threshold calculation unit 120d calculates the background noise value in a plurality of variations. Below, the processes 1-5 which calculate a background noise value are demonstrated in order. The threshold value calculation unit 120d outputs the background noise value calculated by any one of the processes 1 to 5 to the sound detection unit 120e. In addition, when the threshold value calculation unit 120d executes the processes 1 to 3, the similar sound data group 110 may not be normalized. Further, in the processes 1 to 5, the case where the background noise value is calculated using two partial sound data will be described, but the background noise value can be similarly calculated for three or more partial sound data. it can.

  Processing 1 for calculating the background noise value will be described. The threshold calculation unit 120d compares the sound intensity values of the partial sound data of the similar sound data group 110c, and calculates the maximum sound intensity value among the sound intensity values as the background noise value.

  Processing 2 for calculating the background noise value will be described. The threshold calculation unit 120d compares the sound intensity values of the partial sound data of the similar sound data group 110c, and calculates the minimum sound intensity value among the sound intensity values as the background noise value.

  Processing 3 for calculating the background noise value will be described. The threshold calculation unit 120d refers to the sound intensity value of each partial sound data of the similar sound data group 110c, and calculates the average value of the sound intensity values as the background noise value.

  Processing 4 for calculating the background noise value will be described. The threshold calculation unit 120d calculates an average value for each sampling time interval for the sound intensity value of each partial sound data of the similar sound data group 110c, and calculates the calculated average value for each sampling time interval as a background noise value. To do.

  FIG. 3 is a diagram for explaining the process 4 for calculating the background noise value. Here, as an example, the process of the threshold value calculation unit 120d will be described using the partial sound data 10a and the partial sound data 10b3.

  The sound intensity value A1 of the partial sound data 10a is 5, the sound intensity value A2 is 5, the sound intensity value A3 is 4, the sound intensity value A4 is 3, the sound intensity value A5 is 4, and the sound intensity value A6 is 5 The sound intensity value B1 of the partial sound data 10b3 is 3, the sound intensity value P1 is 5, the sound intensity value P2 is 5, the sound intensity value P3 is 4, the sound intensity value P4 is 4, and the sound intensity value B4 is 5

  The time of the sound intensity values A1 and B1 is T1, the time of the sound intensity values A2 and P1 is T2, and the time of the sound intensity values A3 and P2 is T3. The time of sound intensity values A4 and P3 is T4, the time of sound intensity values A5 and P4 is T5, and the time of sound intensity values A6 and B4 is T6.

  The threshold calculation unit 120d calculates an average value of sound intensity values for the same time, and generates a background noise value 20a. The background noise value 20a includes C1 to C6. C1 is an average value of the sound intensity value A1 and the sound intensity value B1 at time T1. C2 is an average value of the sound intensity value A2 and the sound intensity value P1 at time T2. C3 is an average value of the sound intensity value A3 and the sound intensity value P2 at time T3. C4 is an average value of the sound intensity value A4 and the sound intensity value P3 at time T4. C5 is an average value of the sound intensity value A5 and the sound intensity value P4 at time T5. C6 is an average value of the sound intensity value A6 and the sound intensity value B4 at time T6.

  In addition, in the process 4, for the sake of simplicity, the example in which the threshold value calculation unit 120d calculates the average value of each time is shown, but the present invention is not limited to this. The threshold value calculation unit 120d may set the maximum value among the sound intensities at each time to the sound intensity values C1 to C6 of the background noise value 20a, or set the minimum value to the sound intensity values C1 to C6 of the background noise value 20a. It may be set to.

  Processing 5 for calculating the background noise value will be described. The threshold value calculation unit 120d calculates the average value of the sound intensity values of the partial sound data of the similar sound data group 110c at predetermined time intervals, and calculates each calculated average value as the background noise value.

  FIG. 4 is a diagram for explaining the process 5 for calculating the background noise value. Here, as an example, the process of the threshold value calculation unit 120d will be described using the partial sound data 10a and the partial sound data 10b3. In FIG. 4, the description regarding the partial sound data 10a and 10b3 is the same as the description regarding the partial sound data 10a and 10b3 in FIG.

  The threshold calculation unit 120d generates the background noise value 20b by calculating the average value of the sound intensity values included in the predetermined time interval. The background noise value 20b includes sound intensity values D1 and D2. The sound intensity value D1 is an average value of the sound intensity values A1 to A3, B1, P1, and P2 included in the times T1 to T3. The sound intensity value D2 is an average value of the sound intensity values A4 to A6, P3, P4, and B4 included in the times T4 to T6.

  In addition, for the sake of simplicity in the processing 5, the example in which the threshold value calculation unit 120d calculates an average value in a predetermined time period has been shown, but the present invention is not limited to this. The threshold value calculation unit 120d may set the sound intensity values D1 and D2 of the background noise value 20b to the maximum value among the sound intensities in a predetermined time period, or set the sound intensity value D1 of the background noise value 20b to the minimum value. , D2 may be set.

  By the way, the threshold value calculation unit 120d may obtain the sound level value of each frequency instead of the sound intensity value, or may use a value obtained by applying an arbitrary spatiotemporal filter to the partial sound data. For example, the threshold value calculation unit 120d may apply a filter that cuts high-frequency components for noise removal, or applies a filter that emphasizes frequency components similar to the vibration waveform in accordance with the vibration waveform of the vehicle. Also good.

  A case where a filter that emphasizes a frequency component similar to a shaking waveform is applied will be described. The threshold value calculation unit 120d, for example, can extract background noise that is closely related to the shaking of the vehicle by using a filter that emphasizes the sound response repeated in small increments when there is a lot of small shaking.

  In addition, when the similar sound data group 110c does not include a sufficient number of partial sound data, the threshold calculation unit 120d outputs data indicating that the background noise value is not calculated to the sound detection unit 120e. The minimum number of partial sound data used by the threshold calculation unit 120d is appropriately set by the user.

  The sound detection unit 120e is a processing unit that detects a sound to be detected based on the background noise value and the sound data 110a. The sound detection unit 120e may output the detection result to an external device or display it on the display unit 107.

  As an example, the sound detection unit 120e according to the first embodiment detects an abnormal sound such as a horn, but is not limited thereto. Two types of processing in which the sound detection unit 120e detects abnormal sounds will be described. The sound detection unit 120e may detect an abnormal sound using any process.

  The first process of the sound detection unit 120e will be described. The sound detection unit 120e compares the sound intensity value at each time of the sound data with the background noise value. The sound detection unit 120e detects sound data in a time zone in which the sound intensity value is greater than the background noise value and the difference between the sound intensity value and the background noise value is equal to or greater than a predetermined threshold as an abnormal sound.

  Next, the second process of the sound detection unit 120e will be described. The sound detection unit 120e first detects an abnormal sound from the sound data 110a in the same manner as in the prior art. For example, the sound detection unit 120e performs pattern matching between the waveform indicating the characteristics of the abnormal sound and the sound data 110a, and determines the sound data in a time zone in which the similarity is equal to or greater than a predetermined threshold as an abnormal sound. Alternatively, the sound detection unit 120e may compare a predetermined threshold value with the sound intensity value of the sound data 110a and determine sound data in a time period exceeding the predetermined threshold value as an abnormal sound. In addition, the sound detection unit 120e may detect an abnormal sound by any method.

  The sound detection unit 120e corrects the detection result using the background noise value after detecting the abnormal sound. Specifically, the sound detection unit 120e compares the sound intensity value of each abnormal time zone with the background noise value, and detects the abnormal sound in the time zone having a sound intensity value smaller than the background noise value. Is determined. The sound detection unit 120e detects the final abnormal sound by deleting the noise from the abnormal sound.

  Note that the sound detection unit 120e compares the addition value obtained by adding a predetermined threshold to the background noise value and the sound intensity in each time zone of the abnormal sound, and detects the abnormal sound in the time zone in which the sound intensity is lower than the addition value. It may be determined as noise.

  In addition, when the sound detection unit 120e acquires data indicating that the background noise value is not calculated from the threshold value calculation unit 120d, the sound detection unit 120e detects an abnormal sound from the sound data 110a as in the related art. In addition, the sound detection unit 120e may detect an abnormal sound from the sound data 110a based on the accompanying data 110b even when there is no vehicle body shake that exceeds the threshold value.

  Next, an example of a processing procedure of the sound detection apparatus 100 according to the first embodiment will be described. FIG. 5 is a flowchart illustrating the processing procedure of the sound detection apparatus according to the first embodiment. The process illustrated in FIG. 5 is executed, for example, when an abnormal sound detection request is received from the input unit 106 or the like.

  As shown in FIG. 5, the sound detection device 100 determines whether or not there is a vehicle body shake that is equal to or greater than a threshold value based on the accompanying data 110b (step S101). The sound detection device 100 detects an abnormal sound from the sound data when there is no vehicle body shake exceeding the threshold (No in Step S101) (Step S102).

  The sound detection device 100 determines whether or not a predetermined number or more of partial sound data is included in the similar sound data group 110c when the vehicle body shake exceeding the threshold value is present (step S101, Yes) (step S103). ). When the similar sound data group 110c does not include a predetermined number or more of partial sound data (step S103, No), the sound detection device 100 proceeds to step S102.

  When the similar sound data group 110c includes a predetermined number or more of partial sound data (step S103, Yes), the sound detection device 110 normalizes each partial sound data included in the similar sound data group 110c. (Step S104).

  The sound detection device 110 calculates a background noise value (step S105), and detects an abnormal sound based on the background noise value and the sound intensity value of the sound data (step S106).

  Next, the effect of the sound detection apparatus 100 according to the first embodiment will be described. The sound detection device 100 detects a set of partial sound data having similar vehicle swing characteristics from the sound data 110a as a similar sound data group 110c, and based on the similar sound data group 110c, the sound intensity of background noise A background noise value as an index of the value is calculated. And the sound detection part 110 detects the sound data of a detection target from the sound data 110a based on the characteristic of the sound used as a detection target, and corrects a detection result using a background noise value. For this reason, according to the sound detection apparatus 100, it is possible to accurately detect the sound to be extracted even when background noise is included.

  In addition, the sound detection device 100 normalizes each partial sound data included in the similar sound data group 100c. Therefore, the background noise value can be calculated by absorbing the deviation of the sound generation timing of the background noise, and the sound to be detected can be detected with high accuracy.

  In addition, the sound detection device 100 selects a set of partial sound data in a section having similar vehicle shaking characteristics as the similar sound data group 110c by performing DP matching, for example. Therefore, the background noise value can be calculated more accurately, and the detection target sound can be detected with high accuracy.

  Next, a sound detection apparatus according to the second embodiment will be described. For example, it is assumed that the sound detection device is installed on a moving body such as a vehicle. FIG. 6 is a functional block diagram illustrating the configuration of the sound detection apparatus according to the second embodiment. As shown in FIG. 6, the sound detection device 200 includes a microphone 201, an acceleration sensor 202, a speed sensor 203, and a position data acquisition unit 204. The sound detection device 200 includes an interface unit 205, an input unit 206, a display unit 207, a storage unit 210, and a control unit 220.

  The description of the microphone 201, the acceleration sensor 202, the speed sensor 203, and the position data acquisition unit 204 is the same as that of the microphone 101, the acceleration sensor 102, the speed sensor 103, and the position data acquisition unit 104 described in the first embodiment. Further, the interface unit 205, the input unit 206, and the display unit 207 are the same as those described in the interface unit 105, the input unit 106, and the display unit 107 described in the first embodiment.

  The storage unit 210 is a storage unit that stores sound data 210a, accompanying data 210b, a similar sound data group 210c, and a reliability parameter 210d. The storage unit 210 corresponds to, for example, a semiconductor memory device such as a RAM, a ROM, or a flash memory, or a storage device such as a hard disk or an optical disk.

  The description of the sound data 210a, the accompanying data 210b, and the similar sound data group 210c is the same as the description of the sound data 110a, the accompanying data 110b, and the similar sound data group 110c shown in the first embodiment.

  The reliability parameter 210d includes a reliability indicating the likelihood of the abnormal sound detection result. For example, the reliability value ranges from 0 to 1, and the closer to 1, the higher the certainty of the corresponding abnormal sound. For example, the reliability parameter 210d associates reliability with each time segment of abnormal sound.

  The control unit 220 includes a data management unit 220a, a selection unit 220b, a normalization unit 220c, a threshold value calculation unit 220d, and a sound detection unit 220e. The control unit 220 corresponds to an integrated device such as an ASIC or FPGA, for example. The control unit 220 corresponds to an electronic circuit such as a CPU or MPU.

  Among these, the description of the data management unit 220a, the selection unit 220b, the normalization unit 220c, and the threshold value calculation unit 220d is the same as the description of the data management unit 120a, the selection unit 120b, the normalization unit 120c, and the threshold value calculation unit 120d of the first embodiment. It is the same.

  The sound detection unit 220e is a processing unit that detects a detection target sound. The sound detection unit 220e may output the detection result to an external device or display it on the display unit 207. In the second embodiment, as an example, the detection target sound is an abnormal sound.

  Specifically, the processing of the sound detection unit 220e will be described. First, the sound detection unit 220e detects an abnormal sound from the sound data 210a in the same manner as in the prior art. Subsequently, the sound detection unit 220e takes the difference between the sound intensity value of the abnormal sound at each time and the background noise value, and generates a reliability parameter 210d.

  A process in which the sound detection unit 220e generates the reliability parameter 210d will be described. FIG. 7 is a diagram for explaining processing for generating a reliability parameter. In FIG. 7, data 30 is data of abnormal sound detected by the sound detection unit 220e using the conventional technique. The abnormal sound data 30 includes sound intensity values E1 to E6. Let t be the time interval between the sound intensity values. In FIG. 7, data 20b is a background noise value. The background noise value 20b corresponds to the background noise value 20b shown in FIG.

  The sound detection unit 220e obtains a difference between the sound intensity value of the abnormal sound data 30 and the sound intensity value of the background noise value 20b at each time, and calculates a value obtained by multiplying a predetermined correction value as the reliability. For example, the difference between the sound intensity value E1 and the sound intensity value D1 is 6.5, the difference between the sound intensity value E2 and the sound intensity value D1 is 6.5, and the difference between the sound intensity value E3 and the sound intensity value D1 Is set to 6.5. Further, the difference between the sound intensity value E4 and the sound intensity value D2 is 2.2, the difference between the sound intensity value E5 and the sound intensity value D2 is 2.2, and the difference between the sound intensity value E6 and the sound intensity value D2. Is 2.2.

  When the value of the correction value is 0.1, the reliability at time T1 to T3 is 0.65, and the reliability at time T4 to T6 is 0.22. The sound detection unit 220e causes the storage unit 210 to store the obtained reliability of each time as the reliability parameter 210d.

  Note that the sound detection unit 220e sets the reliability for the abnormal sounds T1 to T6 to 1 when data indicating that the background noise value is not calculated from the threshold value calculation unit 220d.

  After generating the reliability parameter 210d, the sound detection unit 220e sorts out abnormal sounds based on the value of the reliability parameter 210d and the threshold value. Specifically, the sound detection unit 220e does not treat an abnormal sound in a time zone with a reliability lower than the threshold as an abnormal sound, and an abnormal sound in a time zone with a higher reliability than the threshold as a final abnormal sound. To detect.

  The process in which the sound detection unit 220e detects the final abnormal sound will be described with reference to FIG. Here, as an example, the threshold is set to 5. Referring to FIG. 7, the reliability of the times T <b> 1 to T <b> 3 is greater than the threshold value 5. Further, the reliability of the times T4 to T6 is smaller than the threshold value 5. For this reason, the sound detection unit 220e finally detects abnormal sounds from time T1 to time T3 among the abnormal sounds 30 detected by the prior art as abnormal sound data.

  Next, an example of a processing procedure of the sound detection apparatus 200 according to the second embodiment will be described. FIG. 8 is a flowchart illustrating the processing procedure of the sound detection apparatus according to the second embodiment. The process illustrated in FIG. 8 is executed when an abnormal sound detection request is received from the input unit 206 or the like, for example.

  As shown in FIG. 8, the sound detection apparatus 200 detects an abnormal sound from the sound data 210a (step S201), and determines whether or not a predetermined number or more of partial sound data is included in the similar sound data group 210c. (Step S202).

  When the similar sound data group 210c does not include a predetermined number or more of partial sound data (No in step S202), the sound detection device 200 sets each reliability of the reliability parameter 210d to 1 (step S203). ), The process proceeds to step S207.

  On the other hand, when the similar sound data group 210c includes a predetermined number or more of partial sound data (step S202, Yes), the sound detection device 200 properly converts each partial sound data included in the similar data group 210c. (Step S204).

  The sound detection apparatus 200 calculates a background noise value (step S205), and sets each reliability of the reliability parameter 210d based on the difference between the background noise value and the sound intensity value of the abnormal sound (step S206). ). The sound detection apparatus 200 selects abnormal sounds based on the reliability parameter 210d (step S207).

  Next, the effect of the sound detection apparatus 200 according to the second embodiment will be described. While detecting the abnormal sound, the sound detection device 200 calculates a background noise value as an index of the sound intensity value of the background noise based on the similar sound data group 210c, and the difference between the abnormal sound and the background noise value. Based on the above, the reliability parameter 210d for the abnormal sound is calculated. Then, the sound detection device 200 removes a portion corresponding to the background noise from the abnormal sound based on the reliability parameter 210d, and detects the final abnormal sound. For this reason, according to the sound detection apparatus 200, even when the abnormal sound is detected using the conventional technique, the background noise can be removed from the abnormal sound.

  Next, a sound detection apparatus according to the third embodiment will be described. For example, it is assumed that the sound detection device is installed in a moving body such as a vehicle. FIG. 9 is a functional block diagram illustrating the configuration of the sound detection apparatus according to the third embodiment. As illustrated in FIG. 9, the sound detection device 300 includes a microphone 301, an acceleration sensor 302, a speed sensor 303, and a position data acquisition unit 304. The sound detection apparatus 300 includes an interface unit 305, an input unit 306, a display unit 307, a storage unit 310, and a control unit 320.

  The description of the microphone 301, the acceleration sensor 302, the speed sensor 303, and the position data acquisition unit 304 is the same as that of the microphone 101, the acceleration sensor 102, the speed sensor 103, and the position data acquisition unit 104 described in the first embodiment. Further, the interface unit 305, the input unit 306, and the display unit 307 are the same as those described in the interface unit 105, the input unit 106, and the display unit 107 described in the first embodiment.

  The storage unit 310 is a storage unit that stores sound data 310a, accompanying data 310b, similar sound data group 310c, reference sound data 310d, and abnormal sound occurrence probability data 310e. The storage unit 310 corresponds to, for example, a semiconductor memory device such as a RAM, a ROM, or a flash memory, or a storage device such as a hard disk or an optical disk.

  The description of the sound data 310a, the accompanying data 310b, and the similar sound data group 310c is the same as the description of the sound data 110a, the accompanying data 110b, and the similar sound data group 110c shown in the first embodiment.

  The reference sound data 310d is data including the characteristics of a reference abnormal sound. For example, the reference sound data 310d has a relationship between time and sound intensity value of a reference abnormal sound. Note that the reference sound data 310d may have a sound level or the like of each frequency of a reference abnormal sound. The reference sound data 310d is prepared in advance by an administrator.

  The abnormal sound occurrence probability data 310e is information in which the probability of an abnormal sound is associated with each abnormal sound time. FIG. 10 is a diagram illustrating an example of a data structure of abnormal sound occurrence probability data. As shown in FIG. 10, the abnormal sound occurrence probability data 310e associates time with the abnormal sound occurrence probability. For example, in the example shown in FIG. 10, the abnormal sound occurrence probability of the abnormal sound at time T1 is 7.1. The higher the probability of occurrence of this abnormal sound, the higher the possibility that it is an abnormal sound. In contrast, the lower the probability of abnormal sound occurrence, the higher the possibility of background noise.

  The control unit 320 includes a data management unit 320a, a selection unit 320b, a normalization unit 320c, a threshold value calculation unit 320d, and a sound detection unit 320e. The control unit 320 corresponds to an integrated device such as an ASIC or FPGA, for example. Moreover, the control part 320 respond | corresponds to electronic circuits, such as CPU and MPU, for example.

  Among these, the description of the data management unit 320a, the selection unit 320b, the normalization unit 320c, and the threshold value calculation unit 320d is the same as the description of the data management unit 120a, the selection unit 120b, the normalization unit 120c, and the threshold value calculation unit 120d of the first embodiment. It is the same.

  The sound detection unit 320e is a processing unit that detects a detection target sound. The sound detection unit 320e may output the detection result to an external device or display it on the display unit 307. In the third embodiment, as an example, the detection target sound is an abnormal sound.

  Specifically, the processing of the sound detection unit 320e will be described. First, the sound detection unit 320e detects an abnormal sound from the sound data 310a in the same manner as in the prior art. Subsequently, the sound detection unit 320e compares the detected abnormal sound with the reference sound data 310d to generate abnormal sound occurrence probability data 310e.

  An example of processing for calculating the abnormal sound occurrence probability data 310e will be described. The sound detection unit 320e compares the abnormal sound as the detection result with the reference sound data 310d to calculate the similarity. The process for calculating the degree of similarity may use conventional techniques. The sound detection unit 320e registers the similarity value at each time as the abnormal sound occurrence probability in the abnormal sound occurrence probability data 310e. Note that the sound detection unit 320e may calculate the abnormal sound occurrence probability by multiplying the similarity value by a predetermined correction value.

  After calculating the abnormal sound occurrence probability data 310e, the sound detection unit 320e updates the abnormal sound occurrence probability data 310e using the background noise value. Specifically, the sound detection unit 320e obtains the absolute value of the difference between the abnormal sound occurrence probability of the abnormal sound occurrence probability data 310e and the background noise value for each time. Then, the sound detection unit 320e updates the abnormal sound occurrence probability data 310e with the absolute value of the difference between the times as the new abnormal sound occurrence probability at each time.

  After updating the abnormal sound occurrence probability data 310e, the sound detection unit 320e sorts out abnormal sounds based on the abnormal sound occurrence probability at each time. The sound detection unit 320e detects an abnormal sound at a time when the abnormal sound occurrence probability is equal to or higher than a predetermined threshold as a final abnormal sound. The sound detection unit 320e does not treat an abnormal sound at a time when the abnormal sound occurrence probability is less than a predetermined threshold as an abnormal sound.

  Next, an example of a processing procedure of the sound detection apparatus 300 according to the third embodiment will be described. FIG. 11 is a flowchart illustrating the processing procedure of the sound detection apparatus according to the third embodiment. The process illustrated in FIG. 11 is executed when an abnormal sound detection request is received from the input unit 306 or the like, for example.

  As shown in FIG. 11, the sound detection device 300 detects an abnormal sound from the sound data 310a (step S301), and generates an abnormal sound at each time of the abnormal sound from the similarity between the reference sound data 310d and the abnormal sound. The probability is calculated (step S302).

  The sound detection apparatus 300 determines whether or not the similar sound data group 310c includes a predetermined number or more of partial sound data (step S303). When the similar sound data group 310c does not include a predetermined number or more of partial sound data (step S303, No), the sound detection device 300 proceeds to step S307.

  When the similar sound data group 310c includes a predetermined number or more of partial sound data (step S303, Yes), the sound detection device 300 normalizes each partial sound data included in the similar sound data group 310c. (Step S304).

  The sound detection device 300 calculates the background noise value (step S305), calculates the absolute value of the difference between the background noise value and the abnormal sound occurrence probability as a new abnormal sound occurrence probability, and uses the abnormal sound occurrence probability data 310e. Update (step S306). The sound detection device 300 detects the final abnormal sound by comparing the abnormal sound occurrence probability with the threshold value (step S307).

  Next, effects of the sound detection device 300 according to the third embodiment will be described. The sound detection device 300 detects abnormal sound and generates abnormal sound occurrence probability data 310e by comparing the detected abnormal sound with the reference sound data 310d. In addition, the sound detection device 300 calculates a background noise value, updates the abnormal sound occurrence probability data 310e using the background noise value, and detects an abnormal detection result based on the updated abnormal sound occurrence probability data 310e. Select sound. Therefore, according to the sound detection device 300, when the similar sound data group 310c is normalized, an abnormal sound can be detected with high accuracy even when the sound is altered.

  Next, a sound detection apparatus according to the fourth embodiment will be described. For example, it is assumed that the sound detection device is installed on a moving body such as a vehicle. FIG. 12 is a functional block diagram illustrating the configuration of the sound detection apparatus according to the fourth embodiment. As shown in FIG. 12, the sound detection device 400 includes microphones 401a and 401b, an acceleration sensor 402, a speed sensor 403, and a position data acquisition unit 404. The sound detection apparatus 400 includes an interface unit 405, an input unit 406, a display unit 407, a storage unit 410, and a control unit 420.

  The microphones 401a and 401b are microphones that collect ambient sounds. The microphones 401 a and 401 b output sound data to the control unit 420. For example, the sound data includes the intensity of sound data for each predetermined time. The microphones 401a and 401b are arranged at positions separated by a predetermined distance.

  The descriptions of the acceleration sensor 402, the speed sensor 403, the position data acquisition unit 404, and the interface unit 405 are the same as the descriptions of the acceleration sensor 102, the speed sensor 103, the position data acquisition unit 104, and the interface unit 105 of the first embodiment. The description of the input unit 406 and the display unit 407 is the same as the description of the input unit 106 and the display unit 107 in the first embodiment.

  The storage unit 410 stores sound data 410a, accompanying data 410b, and a similar sound data group 410c. The storage unit 410 corresponds to, for example, a semiconductor memory device such as a random access memory (RAM), a read only memory (ROM), or a flash memory, or a storage device such as a hard disk or an optical disk.

  The sound data 410a is data including sound data collected by the microphones 401a and 401b. The explanation of the accompanying data 410b is the same as the explanation of the accompanying data 110b in the first embodiment.

  The similar sound data group 410c includes a plurality of partial sound data of sections in the sound data 410a in which the characteristics of the vehicle shake are similar. The similar sound data 410c is generated by the selection unit 420b described later.

  The control unit 420 includes a data management unit 420a, a selection unit 420b, a normalization unit 420c, a threshold value calculation unit 420d, and a sound detection unit 420e. The control unit 420 corresponds to an integrated device such as an ASIC or FPGA, for example. Moreover, the control part 420 respond | corresponds to electronic circuits, such as CPU and MPU, for example.

  The data management unit 420 a is a processing unit that acquires each data from the microphones 401 a and 401 b, the acceleration sensor 402, the speed sensor 403, and the position data acquisition unit 404 and stores the data in the storage unit 410. The data management unit 420a stores the sound data acquired from the microphones 401a and 401b in the storage unit 410 as the sound data 410a. The data management unit 420a causes the storage unit 410 to store acceleration data, speed data, and position data acquired from the acceleration sensor 402, the speed sensor 403, and the position data acquisition unit 404, respectively, as accompanying data 410b.

  The selection unit 420b compares the sound intensity values of the two types of sound data included in the sound data 410a, and specifies the times of the sections with similar vehicle shake characteristics based on the comparison result. Below, the process of the selection part 420b is demonstrated. Here, of the sound data 410a, sound data acquired from the microphone 401a is referred to as sound data 410A, and sound data acquired from the microphone 401b is referred to as sound data 410B.

  The selection unit 420b compares the sound intensity value of the sound data 410A and the sound intensity value of the sound data 410B every time, and the vehicle is shaking for a time when the difference between the sound intensity values is equal to or greater than a predetermined threshold. Specify as time.

  The selection unit 420b extracts partial sound data corresponding to the specified time from the sound data 410a, and stores a set of each partial sound data in the storage unit 410 as a similar sound data group 410c. The selection unit 420b may extract partial sound data from the sound data 410A in the sound data 410a, or may extract partial sound data from the sound data 410B.

  Further, as in the first embodiment, the selection unit 420b specifies the times of sections in which the characteristics of the vehicle shake are similar based on the accompanying data 410b. The selection unit 420b extracts partial sound data corresponding to the specified time from the sound data 410a, and stores them in the storage unit 410 as a similar sound data group 410c.

  Here, an example of other processing of the selection unit 420b will be described. The time of the section in which the characteristics of the vehicle shake specified by the selection unit 420b based on the accompanying data 410b are similar is set as the first time. The selection unit 420b sets the time when the difference between the sound intensity values of the sound data 410A and 410B is equal to or greater than a predetermined threshold as the second time. The selection unit 420b may specify the time when the first time and the second time overlap as the time when the characteristics of the final vehicle shake are similar. The selection unit 420b extracts partial sound data corresponding to the specified time from the sound data 410a and stores it in the storage unit 410 as a similar sound data group 410c.

  The description of the normalization unit 420c, the threshold value calculation unit 420d, and the sound detection unit 420e is the same as the description of the normalization unit 120c, the threshold value calculation unit 120d, and the sound detection unit 120e described in the first embodiment.

  Next, an example of a processing procedure of the sound detection apparatus 400 according to the fourth embodiment will be described. FIG. 13 is a flowchart illustrating the processing procedure of the sound detection apparatus according to the fourth embodiment. The process illustrated in FIG. 13 is executed, for example, when an abnormal sound detection request is received from the input unit 406 or the like.

  As illustrated in FIG. 13, the sound detection device 400 identifies background noise based on the sound intensity values of the left and right microphones, and generates a similar sound data group 410c (step S401). Here, the left and right microphones correspond to the microphones 401a and 401b.

  The sound detection device 400 normalizes each partial sound data included in the similar sound data group 410c (step S402), and calculates a background noise value (step S403). The sound detection device 400 detects an abnormal sound based on the difference between the background noise value and the sound intensity value of the sound data (step S404).

  Next, effects of the sound detection device 400 according to the fourth embodiment will be described. The sound detection device 400 identifies background noise based on the sound data of the microphones 401a and 401b, and excludes detection results of abnormal sound at the same time as the identified background noise. For this reason, it is possible to eliminate background noise that can occur from the very vicinity of either of the microphones 401a and 401b.

  In the first to fourth embodiments, an example of the processing of the sound detection devices 100 to 400 has been described. However, the processing of the sound detection device is not limited to the above processing. Hereinafter, other processes of the sound detection devices 100 to 400 will be described. Here, description will be made using the reference numerals of the sound detection device 100 used in the first embodiment.

  The sound detection unit 120e of the sound detection device 100 may identify background noise using DP matching and detect sound data to be detected. When the sound detection unit 120e uses DP matching, the normalization unit 120c may not normalize the similar sound data group 110c.

  The sound detection unit 120e selects one partial sound data included in the similar sound data group 110c, compares the selected partial sound data with the sound data 110a, and determines the degree of similarity using a DP matching method. To do. The sound detection unit 120e specifies the time of the sound data 110a in which the similarity value is equal to or greater than the threshold, and specifies the sound data corresponding to the specified time as background noise. The sound detection unit 120e removes sound data determined as background noise from the detection result of the abnormal sound. By executing such processing by the sound detection unit 120e, it is possible to accurately detect abnormal sounds while omitting the time for normalization processing.

  Next, an example of a process in which the sound detection apparatus 100 detects an abnormal sound by removing the influence of background noise will be described. FIG. 14 is a diagram for explaining the processing of the sound detection unit. Each horizontal axis in FIG. 14 is a time axis, and the vertical axis is signal intensity.

  50A will be described. Each line segment 50a of 50A corresponds to the accompanying data 110b. 50b corresponds to the sound data 110a. A similar portion of each partial sound data 50b is registered in the similar sound data group 110c as partial sound data.

  50B will be described. When the sound detection device 100 performs normalization on each partial sound data 50b, the result is shown as 50B in FIG.

  50C will be described. As the sound detection apparatus 100 performs summation averaging on each partial sound data 50b, a background noise value 51 is calculated as indicated by 50C in FIG.

  50D will be described. The sound detection device 100 compares the background noise value 51 and the sound data 52, and detects a portion 52a that clearly deviates as an abnormal sound. Note that the sound data 52 corresponds to the sound data 110a.

  Next, an example of a computer that executes a sound detection program that realizes the same function as the display devices 100 to 400 shown in the embodiments will be described. FIG. 15 is a diagram illustrating an example of a computer that executes a sound detection program.

  As illustrated in FIG. 15, the computer 500 includes a CPU 501 that executes various arithmetic processes, an input device 502 that receives data input from a user, and a display 503. The computer 500 includes a reading device 504 that reads a program and the like from a storage medium, and an interface device 505 that exchanges data with another computer via a network. The computer 500 includes various sensors 506. Various sensors 506 correspond to an acceleration sensor and a speed sensor. In addition, the computer 500 includes a position acquisition device 507 and a microphone 508. The position acquisition device 507 acquires position data using the GPS function. The microphone 508 is a microphone that acquires sound data. The computer 500 also includes a RAM 509 that temporarily stores various types of information and a hard disk device 510. The devices 501 to 510 are connected to the bus 511.

  The hard disk device 510 includes, for example, a data management program 510a, a selection program 510b, a normalization program 510c, a threshold calculation program 510d, and a sound detection program 510e. The CPU 501 reads each program 510 a to 510 e and develops it in the RAM 509.

  The data management program 510a functions as a data management process 509a. The selection program 510b functions as a selection process 509b. The normalization program 510c functions as a normalization process 509c. The threshold calculation program 510d functions as a threshold calculation process 509d. The sound detection program 510e functions as a sound detection process 509e.

  For example, the data management process 509a corresponds to the data management unit 120a. The selection process 509b corresponds to the selection unit 120b. The normalization process 509c corresponds to the normalization unit 120c. The threshold value calculation process 509d corresponds to the threshold value calculation unit 120d. The sound detection process 509e corresponds to the sound detection unit 120e.

  Note that the programs 510a to 510e are not necessarily stored in the hard disk device 510 from the beginning. For example, each program is stored in a “portable physical medium” such as a flexible disk (FD), a CD-ROM, a DVD disk, a magneto-optical disk, and an IC card inserted into the computer 500. Then, the computer 500 may read and execute the programs 510a to 510e from these. Further, the display devices 200 to 400 shown in the other embodiments 2 to 4 are also executed by the computer 500 in the same manner as the sound detection device 100.

  The following supplementary notes are further disclosed with respect to the embodiments including the above examples.

(Appendix 1) Based on sound data acquired from a microphone mounted on a moving body and accompanying data including characteristics of shaking of the moving body while acquiring the sound data, the movement of the sound data A selection unit that selects a set of partial sound data in a section with similar body shaking characteristics as a similar sound data group;
A threshold calculation unit that calculates a threshold based on signal intensity values of a plurality of partial sound data included in the similar sound data group;
A sound detection unit that detects detection sound data indicating detection target sound data from the sound data based on the characteristics of the detection target sound, and corrects the detection sound data based on the detection sound data and the threshold value A sound detection device comprising: and.

(Supplementary Note 2) The threshold value calculation unit calculates an average value, a minimum value, or a maximum value of signal intensity values of each partial sound data of the similar sound data group as a threshold value, and the sound detection unit The sound detection apparatus according to appendix 1, wherein a difference between a signal intensity value and the threshold value is calculated, and the detection sound data is corrected based on the difference.

(Additional remark 3) The said sound detection part calculates the reliability which multiplied the correction value to the difference value of the threshold value and the said detection sound data which the said threshold value calculation part calculated, and calculates the reliability for every time calculated The sound detection device according to appendix 1, wherein the detection sound data is corrected based on the detection sound data.

(Additional remark 4) The said sound detection part is the sound of the detection target in each time of the said detection sound data based on the similarity degree of the reference sound data containing the characteristic of the said detection target sound, and the said detection sound data Calculating an occurrence probability indicating the certainty of the occurrence, updating the occurrence probability based on a difference between the occurrence probability and the threshold value, and correcting the detection sound data based on the updated occurrence probability. The sound detection apparatus according to Supplementary Note 1.

(Additional remark 5) Normalization which normalizes the said similar sound data group by unifying the length of the partial sound data contained in the said similar sound data group, and unifying the space | interval of the sound signal strength value in this partial sound data The sound detection device according to any one of supplementary notes 1 to 4, further comprising a unit.

(Additional remark 6) The said sound detection part specifies the sound data used as background noise by performing DP matching with the partial sound data contained in the said similar sound data group, and the sound data acquired from the said microphone, and specifies The sound detection apparatus according to appendix 1, wherein the detection sound data is corrected based on the result.

(Supplementary note 7) A plurality of the microphones are arranged at different positions of the moving body, and the sound detection unit is from a section other than a section where the signal intensity value of the sound data of any one of the plurality of microphones is equal to or greater than a threshold value. The sound detection apparatus according to appendix 1, wherein detected sound data is detected.

(Appendix 8) A sound detection method executed by a computer,
Based on the sound data acquired from the microphone mounted on the moving body and the accompanying data including the characteristics of the shaking of the moving body during the acquisition of the sound data, among the sound data, the shaking of the moving body Select a set of partial sound data of sections with similar features as a similar sound data group,
Based on signal intensity values of a plurality of partial sound data included in the similar sound data group, a threshold value is calculated,
Detecting detected sound data indicating the detected sound data from the sound data based on the characteristics of the detected sound, and correcting each detected sound data based on the detected sound data and the threshold A sound detection method that is performed.

(Additional remark 9) The process which calculates the said threshold value calculates the average value of the signal strength value of each partial sound data of the said similar sound data group, the minimum value, or the maximum value as a threshold value, and the process which corrects the said detected sound data is The sound detection method according to appendix 8, wherein a difference between a signal intensity value of the detected sound data and the threshold value is calculated, and the detected sound data is corrected based on the difference.

(Additional remark 10) The process which corrects the said detection sound data calculates the reliability which multiplied the correction value to the difference value of the said threshold value and the said detection sound data for every time, and based on the calculated reliability for every time The sound detection method according to appendix 8, wherein the detected sound data is corrected.

(Additional remark 11) The process which corrects the said detection sound data is the detection target in each time of the said detection sound data based on the similarity degree of the reference sound data containing the characteristic of the said detection target sound, and the said detection sound data The occurrence probability indicating the probability of being a sound of the sound is calculated, the occurrence probability is updated based on the difference between the occurrence probability and the threshold value, and the detected sound data is corrected based on the updated occurrence probability. The sound detection method according to appendix 8, wherein:

(Additional remark 12) The process which normalizes the said similar sound data group by unifying the length of the partial sound data contained in the said similar sound data group, and unifying the space | interval of the sound signal strength value in this partial sound data The sound detection method according to any one of appendices 8 to 11, further executed.

(Additional remark 13) The process which corrects the said detection sound data carries out DP matching with the partial sound data contained in the said similar sound data group, and the sound data acquired from the said microphone, The sound data used as background noise is obtained. The sound detection method according to appendix 8, wherein the detected sound data is corrected based on the specified result.

(Supplementary Note 14) A plurality of the microphones are arranged at different positions on the moving body, and in the process of correcting the detected sound data, the signal intensity value of the sound data of any one of the plurality of microphones is equal to or greater than a threshold value. The sound detection method according to appendix 8, wherein detected sound data is detected from other than a section.

(Supplementary note 15)
Based on the sound data acquired from the microphone mounted on the moving body and the accompanying data including the characteristics of the shaking of the moving body during the acquisition of the sound data, among the sound data, the shaking of the moving body Select a set of partial sound data of sections with similar features as a similar sound data group,
Based on signal intensity values of a plurality of partial sound data included in the similar sound data group, a threshold value is calculated,
Detecting detected sound data indicating the detected sound data from the sound data based on the characteristics of the detected sound, and correcting each detected sound data based on the detected sound data and the threshold A sound detection program that is executed.

(Additional remark 16) The process which calculates the said threshold value calculates the average value of the signal strength value of each partial sound data of the said similar sound data group, the minimum value, or the maximum value as a threshold value, and the process which corrects the said detection sound data is The sound detection program according to appendix 15, wherein a difference between a signal intensity value of the detected sound data and the threshold value is calculated, and the detected sound data is corrected based on the difference.

(Additional remark 17) The process which corrects the said detection sound data calculates the reliability which multiplied the correction value to the difference value of the said threshold value and the said detection sound data for every time, and based on the calculated reliability for every time The sound detection program according to appendix 15, wherein the detection sound data is corrected.

(Supplementary Note 18) The process of correcting the detected sound data is performed by detecting the detection target at each time of the detected sound data based on the similarity between the reference sound data including the characteristics of the detection target sound and the detected sound data. The occurrence probability indicating the probability of being a sound of the sound is calculated, the occurrence probability is updated based on the difference between the occurrence probability and the threshold value, and the detected sound data is corrected based on the updated occurrence probability. The sound detection program according to supplementary note 15, characterized in that:

(Additional remark 19) The process which normalizes the said similar sound data group by unifying the length of the partial sound data contained in the said similar sound data group, and unifying the space | interval of the sound signal strength value in this partial sound data. The sound detection program according to any one of supplementary notes 15 to 18, further causing a computer to execute the program.

(Additional remark 20) The process which corrects the said detection sound data carries out DP matching with the partial sound data contained in the said similar sound data group, and the sound data acquired from the said microphone, The sound data used as background noise is obtained. The sound detection program according to appendix 15, wherein the sound detection data is specified and the detected sound data is corrected based on the specified result.

(Additional remark 21) A plurality of the microphones are arranged at different positions of the moving body, and in the process of correcting the detected sound data, the signal intensity value of the sound data of any one of the plurality of microphones is equal to or greater than a threshold value. The sound detection program according to appendix 15, wherein detected sound data is detected from other than a section.

100, 200, 300, 400 Sound detection device 101, 201, 301, 401a, 401b Microphone 102, 202, 302, 402 Acceleration sensor 103, 203, 303, 403 Speed sensor 104, 204, 304, 404 Position data acquisition unit 105 205, 305, 405 Interface unit 106, 206, 306, 406 Input unit 107, 207, 307, 407 Display unit 110, 210, 310, 410 Storage unit 120, 220, 320, 420 Control unit

Claims (9)

Based on the sound data acquired from the microphone mounted on the moving body and the accompanying data including the characteristics of the shaking of the moving body during the acquisition of the sound data, among the sound data, the shaking of the moving body A selection unit that selects a set of partial sound data of sections with similar characteristics as a similar sound data group;
A threshold calculation unit that calculates a threshold based on signal intensity values of a plurality of partial sound data included in the similar sound data group;
A sound detection unit that detects detection sound data indicating detection target sound data from the sound data based on the characteristics of the detection target sound, and corrects the detection sound data based on the detection sound data and the threshold value A sound detection device comprising: and.   The threshold value calculation unit calculates an average value, a minimum value, or a maximum value of signal intensity values of each partial sound data of the similar sound data group as a threshold value, and the sound detection unit calculates a signal intensity value of the detected sound data and The sound detection apparatus according to claim 1, wherein a difference from the threshold value is calculated, and the detected sound data is corrected based on the difference.   The sound detection unit calculates a reliability obtained by multiplying a difference value between the threshold calculated by the threshold calculation unit and the detected sound data for each time, and based on the calculated reliability for each time, The sound detection apparatus according to claim 1, wherein the sound detection data is corrected.   The sound detection unit is, based on the similarity between the reference sound data including the characteristics of the detection target sound and the detection sound data, the probability of being the detection target sound at each time of the detection sound data. The occurrence probability indicating the occurrence probability is calculated, the occurrence probability is updated based on a difference between the occurrence probability and the threshold value, and the detected sound data is corrected based on the updated occurrence probability. The sound detection apparatus according to 1.   It further has a normalization unit that normalizes the similar sound data group by unifying the lengths of the partial sound data included in the similar sound data group and unifying the intervals of the sound signal intensity values in the partial sound data. The sound detection device according to claim 1, wherein the sound detection device is a sound detector.   The sound detection unit performs DP matching on the partial sound data included in the similar sound data group and the sound data acquired from the microphone, thereby specifying sound data serving as background noise, and based on the specified result. The sound detection device according to claim 1, wherein the detection sound data is corrected.   A plurality of the microphones are arranged at different positions of the moving body, and the sound detection unit receives the detected sound data from other than a section where the signal intensity value of the sound data of any one of the plurality of microphones is equal to or greater than a threshold value. The sound detection device according to claim 1, wherein the sound detection device is detected. A sound detection method executed by a computer,
Based on the sound data acquired from the microphone mounted on the moving body and the accompanying data including the characteristics of the shaking of the moving body during the acquisition of the sound data, among the sound data, the shaking of the moving body Select a set of partial sound data of sections with similar features as a similar sound data group,
Based on signal intensity values of a plurality of partial sound data included in the similar sound data group, a threshold value is calculated,
Detecting detected sound data indicating the detected sound data from the sound data based on the characteristics of the detected sound, and correcting each detected sound data based on the detected sound data and the threshold A sound detection method that is performed. On the computer,
Based on the sound data acquired from the microphone mounted on the moving body and the accompanying data including the characteristics of the shaking of the moving body during the acquisition of the sound data, among the sound data, the shaking of the moving body Select a set of partial sound data of sections with similar features as a similar sound data group,
Based on signal intensity values of a plurality of partial sound data included in the similar sound data group, a threshold value is calculated,
Detecting detected sound data indicating the detected sound data from the sound data based on the characteristics of the detected sound, and correcting each detected sound data based on the detected sound data and the threshold A sound detection program that is executed.

Images