JP6518864B1 - Vehicle sound source direction detection system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To realize time responsiveness according to the type of sound source when detecting the sound source direction of each medium such as a horn sound and an alarm sound of an emergency vehicle as a sensor applied for driving assistance and automatic driving of a car Sound source direction detection under the environment where the emergency vehicle alarm sound of this is intercepted by the building and it is out of sight becomes an issue. SOLUTION: A plurality of microphones are arranged inside one section of the upper surface of a vehicle to form directivity in a plurality of directions, thereby decomposing it into a signal for each sound wave incoming direction. Further, the signal for each medium to be detected is separated into an arrival-specific signal and a medium-specific signal by a filter that passes the signal. In addition to detecting the amplitude difference of this signal, the sound source direction is comprehensively detected using a means for detecting the advance / delay of the frequency change pattern of the sound wave from each media in each direction. As a result, even in the case of a single sound like a horn sound, the sound source direction can be correctly detected even with a sound reflected from a distant sound like an emergency vehicle sound or a sound wave blocked by a building or the like. [Selected figure] Figure 1

Description

The present invention relates to a method of detecting a sound source direction of a horn sound around a vehicle and an alarm sound of an emergency vehicle.

In automatic driving, image recognition, laser radar sensing, millimeter wave sensing, etc. are used as sensors, but sound information also includes other information, and sound source type and sound source direction sensing are also necessary sensors including driving assistance Become. The object of the present invention is to detect the sound source direction of a single short-distance sound such as a horn sound, and a diffracted wave and a reflected wave to which a sound wave from a direct direction such as a siren sound of an emergency vehicle attenuates and comes It is to perform sound source direction detection in the environment.
The application field of the conventional sound wave arrival direction detection is an application to an environment where there is a reflector such as a side wall but there is no object that shields the sound wave between the sound source position and the sound source direction detection unit. . Therefore, the sound wave arrival direction detection method has been focused on reducing the detection error due to the influence of the reflected wave. Also, the detection target has a large number of frequency components such as voice, and is a continuous sound to some extent, and detection of a single sound such as a horn sound or an emergency vehicle alarm sound or a sound of a specific frequency is considered. Not.

(Japanese Patent Application Laid-Open No. 2001-166025) shows that the detection angle accuracy is improved by frequency-resolving the sound wave arrival time difference (phase difference) with a plurality of microphones to obtain the arrival angle and combining the detection results. There is no indication of malfunction avoidance measures for waves. (Japanese Patent Application Laid-Open No. 2017-128075) In order to prevent the decrease in accuracy due to the reflected wave when following the front of the system in the sound source direction, a microphone system having directivity formed is used, but the point that rotation control is premised is different. . Also, no measures are taken against the diffracted waves. In JP-A-2017-67666, a microphone array is used to sequentially change the direction of directivity, and the maximum amplitude direction is regarded as the sound source direction, but reflection and attenuation due to an obstacle occur in the sound wave propagation path. The case is not shown. (Japanese Patent Application Laid-Open No. 2010-105357) aims to improve the quality of the recorded sound when acquiring the traveling sound of the proximity vehicle of the own vehicle using a plurality of microphones. Therefore, although the source direction estimation method used uses the time difference of a plurality of frequency components, no countermeasure is taken for the case where reflection or attenuation occurs due to an obstacle during sound wave propagation at a distant source.

In automatic driving, the position of another car is detected by image recognition, laser radar sensing, millimeter wave sensing, etc., but if there is a large car between the target cars to be detected or if it is in the shadow of a building It can not be detected by light or radio waves. In addition, the collision sound emitted by other vehicles also includes warning information and message information such as "Please get rid of it", and includes information that is not included in the sensing of light and radio waves. Furthermore, when an emergency vehicle approaches, a general vehicle is obliged to evacuate, and it is necessary to detect the true direction of the vehicle, that is, the true sound source direction even when the emergency vehicle is positioned out of sight, such as a large vehicle or a building shade.

  In the case of the detection of the horn sound, it is a single sound with a distance of several meters, and it is necessary to detect the sound source direction accurately in an instant. In the case of emergency vehicle alarm sound detection, it is also necessary to detect a sound several tens of meters away, but in this case it may be a shadow of a building or a large vehicle. The sound wave from the sound source also comes as a diffracted sound from the direction of the sound source, but the amplitude attenuates. On the other hand, the sound wave repeatedly reflected from the sound source by a building etc. also attenuates. The sound wave arrival direction in the direction is not limited to the sound source position, that is, the position of the emergency vehicle.

  The problem is to detect the sound source direction correctly for such a situation. Furthermore, it is also an issue how and at what position of the car the microphone is mounted. For detecting the direction of arrival of sound waves, directivity formation with a plurality of microphones is used, but the realization of a structure in which a time difference corresponding to the angle of arrival of the sound waves necessary for forming directivity is obtained is an issue.

  In the method of the present invention, a plurality of microphones are arranged in one section facing the external space of the vehicle, and directivity of a plurality of directions is formed by arithmetic synthesis of the microphone output signals, thereby obtaining signals for each sound wave arrival direction obtained Disassemble. From the output, a signal for each medium to be detected, such as a horn sound and an emergency vehicle alarm sound, is separated into an arrival-specific signal and a media-specific signal by a filter that passes the signal. The source direction is detected from the separated signal after this separation.

  The horn sound is a near-field sound, and since direct sound is dominant and it is necessary to make an instantaneous judgment, the direction with the largest amplitude value among the amplitudes from each direction is regarded as the sound source direction. Furthermore, a method of dividing the amplitude in each direction into an X-direction component and a Y-direction component and obtaining the sound source direction by Arctan calculation which is an inverse tangent function is also a method of enhancing resolution.

  It is necessary for the emergency vehicle to detect the sound source direction of the emergency vehicle from the state of being located at a distance. In that case, a diffraction wave and a reflected wave are generated due to the presence of an object such as a large vehicle and a building that shields or reflects the sound wave between the sound source and the sound receiving position. Diffracted waves have large amplitude attenuation and come from the direction of the source substantially, while reflected waves come from a direction different from the direction of the true source, but the amplitude attenuation may be smaller than the diffracted wave. Therefore, if the direction of arrival is determined only by the amplitude information, false detection may occur.

In addition, although a sound wave attenuates by the phenomenon in which a wave goes around behind the obstacle and a wave advances according to the principle of Hoygens with a diffraction wave, the course which propagates linearly also exists.
Therefore, processing is performed to detect temporal advance and delay of the frequency component change pattern of the sound wave from each direction after separation into arrival direction and by medium signals, and regard the direction in which the change pattern is most advanced as the sound source direction. This makes it possible to regard the arrival direction of the diffracted wave that is the correct direction as the sound source direction. In this method, several seconds are required for detection, but the emergency vehicle has a large volume and can detect sounds from a distance, and it is possible to detect a time difference and detect a correct sound source direction over several seconds.

It is necessary to arrange the microphone on the flat surface of the top of the vehicle as the mounting place of the microphone, but as an example of the installation place, one part of the rear spoiler mounted on many vehicles is remodeled and arranged inside There is a way. In this case, a small microphone receiving hole of a size of several mm is made on the hard plastic upper part of the rear spoiler, and the upper part is covered with a soft material which transmits air vibration but does not pass water.

In the microphone system placed at a distance of several tens of centimeters or more in the vehicle, the correlation between the amplitude of the microphone input and the arrival time is reduced, but multiple microphones are arranged in a narrow range of 10 centimeters × 10 centimeters or less By forming the directivity, the correlation between the amplitude of each directivity output and the arrival time becomes high, and the amplitude difference and arrival time difference detection of each directivity direction can be performed accurately.

Even with media that require instantaneous determination, multiple filters can be created by changing the determination method using the same hardware, even if there is diffraction or reflection of sound waves from a distance, by means of a filter that separates the directivity of the microphone and the media to be detected. The sound source direction of the media can be detected correctly.

Further, by combining the directivity formation in a plurality of directions with the temporal advance / delay detection of the frequency component change pattern, it becomes possible to correctly detect the sound source direction of the emergency vehicle sound in which diffracted waves and reflected waves are out of sight.

Overall configuration of the present invention (vehicle sound source direction detection system) Frequency selective filter characteristics corresponding to detection target Example of detecting the sound source direction from the amplitude difference in the arrival direction Diffraction wave and reflection wave propagation example when emergency vehicle is out of sight Sound source direction detection procedure of emergency vehicle alarm sound Lead-lag detection method of frequency component change pattern of emergency vehicle alarm sound Frequency component change pattern lead / lag detection signal processing waveform example Characteristics of frequency component change pattern lead / lag detection output Example of mounting a microphone on the rear spoiler surface Basic configuration of directivity formation method Figure 10: Directional characteristics of the configuration Characteristic example of four directional patterns Characteristic example of eight directional patterns

An embodiment of the overall configuration in the case of four directional patterns is shown in FIG. A plurality of microphones are installed on the upper plane of the vehicle, and finger characteristics in four directions are formed by addition, subtraction, integration, etc., and the output is divided by filters that detect each medium such as horn sound detection, ambulance sound detection, police car sound, etc. Input to the detection processing block for each medium to detect the sound source direction.

  An example of the filter characteristic of the frequency domain for separating and extracting the sound to be detected is shown in FIG. The horn sound is a combination of multiple dissonances of 310 Hz to 580 Hz, and uses a signal that has passed through a band pass filter of 310 Hz to 580 Hz. Among emergency vehicles, an ambulance is a sound repetition of 960 Hz at 0.65 seconds and 770 Hz at 0.65 seconds, and uses signals that pass through 770 Hz and 960 Hz band pass filters. The police car's siren sound is a sweep sound that changes from 430 Hz to 870 Hz in a 4-second cycle or an 8-second cycle, but detected with a band pass filter of 600 Hz to 700 Hz that does not overlap with the horn sound and the ambulance sound. In addition, although the engine sound of a vehicle, a road noise, etc. are mixed, since the sound pressure is high, the sound of the horn and the alarm sound of an emergency vehicle can be suppressed by separation with a filter.

The horn sound detection targets only a short distance sound of about several meters, is single sound generation, and requires instantaneous detection. Therefore, the sound source direction is detected using the amplitude difference after passing through the detection filter in the directivity output signal of the microphone. The directivity direction with the highest amplitude value among the directivity outputs may simply be regarded as the sound source direction, but the angular resolution in that case is limited to the number of directivity directions formed. Therefore, if the amplitude for each directivity is decomposed in the X direction component and the Y direction, and arctangent function calculation (Arctan) of XY amplitude is performed, it can be decomposed into continuous angles. Since this processing does not involve processing in the time domain, detection is possible in about 0.1 sec.

An embodiment in which the arrival direction of the sound wave is accurately determined by the amplitude difference is shown in FIG. The amplitude vector in the X direction is determined from the amplitude difference in the lateral direction, ie, the X direction, and the Y direction vector is determined from the amplitude difference in the longitudinal direction, ie, the Y direction, to increase the resolution in the arrival direction by Arctan (inverse tangent) calculation.
The equation for obtaining the angle is shown in equation 1.

When approaching an emergency vehicle, general vehicles are obliged to evacuate and it is necessary to detect from sounds several tens of meters ahead. On the other hand, the emergency vehicle sound regularly changes the frequency component of the sound in a relatively long cycle (1.3 seconds for ambulance). In the case of detecting a sound from several tens of meters ahead, a diffracted sound and a reflected sound are produced due to the influence of a building or a large vehicle, and the sound source direction can not be detected correctly only with the amplitude information. Therefore, in addition to the amplitude value of the signal from each direction, the time lead-lag of the signal frequency component from each direction is also detected, and when the lead-lag time difference is small, the sound source direction is determined only by the amplitude value, and the time difference of lead-lag is large In this case, processing is performed in which the directional direction most advanced in time is regarded as the true sound source direction.

Fig. 4 shows an example of the diffracted wave and reflected wave of sound wave propagation when the emergency vehicle is out of sight. If a building or the like exists between the vehicle and the emergency vehicle, it attenuates, but the diffracted wave propagates from the direction of the sound source. In addition to this, there is also a sound wave coming from the reflected wave, and the amplitude of the sound wave in that direction may be larger than the diffracted wave. Therefore, the direction of arrival merely having a large amplitude is not limited to the direction of the sound source. Diffraction reduces the amplitude of the propagating sound wave, but the distance is the shortest, resulting in the most advanced signal in the time domain. Therefore, if the signal detects the most advanced directional direction in time, the direction of the emergency vehicle can be correctly detected. When the difference between propagation paths is 10 m, the time difference is 30 mSec. For example, one cycle of 500 Hz is 2 mSec, and the advance / delay can not be determined from the signal phase difference. Therefore, it is necessary to detect the advance / delay of time in advance of the frequency component change pattern of the emergency vehicle alarm sound.

An example of processing when the medium to be detected is an emergency vehicle alarm sound is shown in FIG. First, levels in a plurality of directions are detected, and if one or more levels are levels which are not affected by noise other than the medium to be detected, the direction is regarded as the sound source direction.
When the level above the specified value is in multiple directions, temporal advance in each direction of the change pattern of the frequency component with reference to the amplitude maximum direction, detecting the delay and detecting whether there is a time difference more than the specified value If it is less than the value, the maximum amplitude direction is regarded as the sound source direction. If the time difference between the plurality of directions is equal to or greater than a certain value (the difference is large), the directivity direction most advanced in time is determined as the sound source direction.
The time difference is determined to be more than or equal to about 10 mSec, which is a time difference within several meters, because it depends on whether the time difference is within the range or not.

An embodiment of a method for detecting the advance / delay of the frequency component change pattern of the emergency vehicle alarm sound is shown in FIG. The signal output classified by media and oriented gender is binarized by the SGN function (sign function). By the capture effect of the binarization processing, the frequency component with the largest amplitude can be extracted and the influence of noise can be eliminated. A value proportional to the frequency is generated from the rising and falling time intervals of this binarized signal, and the variation pattern is passed through a band pass filter. This band pass filter is for extracting ones in which the change time of the frequency according to the medium to be detected is within a certain range.

  The signals obtained by subtracting the arrival signals VfA and VfB from the two directions that have passed through the band pass filter are integrated and the signal Vadd obtained by adding Vint, VfA and VfB, as shown in Eqs. 2 to 6, and after averaging The signal Laed is Td, that is, a signal proportional to the arrival time difference between the A direction signal and the B direction signal.

VA: A direction signal amplitude VB: B direction signal amplitude
Td: Arrival time difference between A direction signal and B direction signal
Tper: Frequency change period τ: Integral time constant

An example of the lead / lag detection processing waveform is shown in FIG. 7. As the detection value Laed, positive and negative outputs corresponding to the lead and lag of the A and B signals are obtained. Further, a lead / lag detection characteristic with respect to "signal time difference / signal frequency change period" is shown in FIG. If “signal time difference / signal frequency change period” is ± 0.2 or less, it can be judged correctly, but the frequency change period Td of ambulance alarm sound is 1.3 seconds, 0.2 times of that is 0.26 seconds, and the distance difference is equivalent to 85 m Even time differences can be detected.

The placement of the microphone on the vehicle is required to have no protrusion from the vehicle and to not affect the strength of the vehicle. On the other hand, in order to form the directivity in the horizontal direction in the horizontal direction by the plurality of microphones, it is preferable that the plurality of microphones can be arranged in a plane close to the plane of one stroke of the vehicle. Also, it is required that the incoming sound wave does not become a shadow of the vehicle.
An example mounted on the rear spoiler surface as a position satisfying the condition is shown in FIG. In order to receive sound waves coming from space, a small microphone-sized hole of several mm is made on the upper surface of the rigid rear spoiler, and a soft material that is waterproof but allows air vibration is installed. As a result, the plurality of microphones in one section of the vehicle corresponds to the characteristic disposed almost in free space, and a plurality of directional patterns are formed by calculation of these microphone signals.
The inter-microphone distance can be formed with directivity of about 5 cm with four microphones.

FIG. 10 shows an example of directivity formation by two microphones which is the basis of a method of forming directivity from a plurality of microphones in a plane close to the plane of a section of a vehicle. This characteristic is the operation shown in the equations 7 to 9, and the directivity characteristic is as shown in FIG. By combining such configurations, directivity in a plurality of directions can be formed. In the case of this configuration, it is an arrangement example of the inter-microphone distance 5 Cm.
If two or more basic two pairs are used, four-directional directivity and eight-directional directivity can be formed.
Examples of directivity to be formed are shown in FIG. 12 (example of four directional directivity) and FIG. 13 (example of eight directional directivity).

k = ω / c c: sound velocity
ω: source angular frequency d: distance between microphones

here

If you set

101: A plurality of microphones arranged in one section of the upper surface of the vehicle
102: Multiple directional directivity is formed by composition involving operation
103: A direction directivity output
104: B direction directivity output
105: C direction directivity output
106: D direction directivity output

111: Signal after extraction of horn sound in direction A
112: Signal after extraction of horn sound in direction B
113: Signal after extraction of horn sound in direction C
114: Signal after extraction of horn sound in direction D

131: Signal after police car siren sound extraction of direction A
132: Signal after police car siren sound extraction of direction B
133: Direction C signal after police car siren sound extraction
134: Signal after police car siren sound extraction in direction D

140: Sound source direction detection processing unit of horn sound
141: Source direction detection processing unit for ambulance warning sound
142: Sound source direction detection processing unit of police car siren sound

311: Filtered signal of media to be detected in directional output in the right direction
312: Filtered signal of media to be detected in directional output in left direction
313: Filtered signal of media to be detected in directional output in forward direction
314: Filtered signal of media to be detected in directional output in the backward direction
320: Angle detector by Arctan (inverse tangent) processing

610: A direction directional signal The signal passed through the extraction filter is binarized by the SGN function
611: B-directional signal The signal passed through the extraction filter to the directional signal is binarized by the SGN function
620: Converted to a value proportional to the frequency of the A-direction binarized signal
621: Convert to a value proportional to the frequency of B-direction binarized signal
630: Band limiting filter for A direction frequency change signal
631: Band limiting filter for B direction frequency change signal
640: time integration unit
650: Time averaging unit

910: Equipped with multiple microphones with waterproofing inside the rear spoiler of the vehicle

Claims (6)

In a system for detecting a sound source direction for each medium to be detected, such as a horn sound and an alarm sound of an emergency vehicle, for driving assistance of a car and automatic driving.
A plurality of microphones are arranged in one section within 10 cm × 10 cm of the plane facing the upper space of the vehicle, and combining processing accompanied by calculation of the signals forms directivity in multiple directions, and the directivity output is branched, Sound source direction detection processing is performed from signals separated for each medium for each directivity direction via a frequency selection filter that passes the signal for each medium to be detected,
For media that should be instantaneously detected within a few meters of the distance between the sound source and the vehicle, such as a horn sound, the direction of the maximum amplitude is determined from the separated signals and considered as the sound source direction,
For media in which the distance between the sound source and the own vehicle, such as alarm sounds of emergency vehicles, exceeds a few meters and reflections and obstacles are expected due to sound wave propagation from the obstacles, the above separated signals give rise to the temporal change of frequency component change pattern Detecting lead and delay,
If it is judged that the lead-lag time difference is as small as approximately 10mSec or less, the maximum amplitude direction is determined from the above separated signal and regarded as the sound source direction. If the lead-lag time difference is larger than that, the most time is taken from the separated signal. An on-vehicle sound source direction detection system that performs sound source direction detection processing in which the directivity direction that has been advanced is regarded as the sound source direction.   As a method of obtaining the maximum amplitude direction, a method in which the classification direction in which the maximum amplitude is obtained is regarded as the sound source direction from signals separated into each media direction, or forward directivity amplitude, backward directivity amplitude, left directivity The sound source direction is detected by detecting the dynamic amplitude and the right directional directivity amplitude, decomposing it into the amplitude in the Y direction and the amplitude in the X direction, and setting the sound source direction angle = Arctan (amplitude in Y direction / amplitude in X direction). Vehicle sound source direction detection system.   The vehicle sound source direction according to claim 1, wherein when detecting the change pattern of the frequency component of the directional output in each direction for each medium, the signal is converted to a value proportional to the frequency after being binarized by the code function. Detection system.   When detecting the time lead / delay of the frequency component change pattern of 2 systems, convert the signal proportional to the frequency, then set the 2 system signal passed through the band pass filter as VfA, VfB and subtract VfA, VfB and integrate The vehicle sound source direction detection according to claim 1, wherein Vint is Vint and a signal obtained by adding VfA and VfB is Vadd, and after Vint is multiplied by Vadd, the signal advance / delay is detected depending on whether Laed which is an averaged signal is positive or negative. system. A plurality of holes of several mm equivalent to the microphone receiver are made in one section of the plane facing the upper space of the vehicle, and the upper part is made of a waterproof material that is covered with a soft material through which sound vibration passes. The vehicle-mounted sound source direction detection system according to claim 1, wherein a plurality of microphones are installed under the opened portion. 6. The on-vehicle sound source direction detection system according to claim 5, wherein the plurality of microphones are installed in a rear spoiler of a car.

Images