JP2012058340A - Noise detection apparatus, noise detection method and noise detection system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a noise detection apparatus capable of calculating the cycle of footsteps even if there is fluctuation in walking cycle or partial missing in footstep detection, a noise detection method and a noise detection system.SOLUTION: A noise detection apparatus includes: a noise detection section 102 for detecting a signal of a noise; a cycle calculation section 104 which fluctuates the cycle of a triangular function in such a manner that an inflection point of a waveform drawn by the triangular function is matched to a plurality of detected noises and determines the cycle of the waveform matching the inflection point to the noises as the cycle of noises; and a determination section 105 which determines whether the determined cycle of noises is settled within a predetermined range and, when it is determined that the cycle of noises is settled within the range, determines that the plurality of detected noises are detection target sounds.

Description

  The present invention relates to a sound detection device, a sound detection method, and a sound detection system.

  2. Description of the Related Art Conventionally, there is known a security device that detects an intrusion of a suspicious person in a monitoring area based on an image captured by a monitoring camera, a sensor detection signal, and a sound collected by a microphone. For example, Patent Document 1 describes a method for determining whether or not a footstep sound is captured by capturing the frequency characteristics of a physical sound.

  However, in the method of Patent Document 1, it is difficult to distinguish footsteps from sound caused by falling objects. In addition, since the frequency characteristics of footsteps vary depending on footwear and flooring of pedestrians, there is a problem that a large amount of footstep reference data needs to be prepared in order to detect footsteps.

  Therefore, Patent Document 2 discloses a method for determining footsteps from the periodicity of sound. Patent Document 2 discloses a method of determining a footstep when a sound collected from a microphone and a stored waveform pattern are matched by using a waveform pattern storage unit that stores a plurality of footstep waveform patterns. Has been.

JP 2000-066941 A JP 2006-148357 A

  However, in general, human walking is periodic, but fluctuations occur in walking due to the pedestrian's habit, etc., and thus it may be difficult to calculate the walking period using a pattern that has been patterned in advance.

  The present invention has been made in view of the above, and a sound detection device, a sound detection method, and a sound detection method capable of calculating a footstep cycle even when there are fluctuations in the walking cycle and lack of footstep detection. The object is to provide a sound detection system.

  In order to solve the above-described problems and achieve the object, a sound detection device according to the present invention includes a detection unit that detects a plurality of sound signals that are continuous from an input sound, and the detected sound. A period calculation unit that varies the period of the trigonometric function so that the inflection point of the waveform drawn by the trigonometric function matches, and determines the period of the waveform at which the inflection point matches the physical sound as the period of the physical sound; Determining whether or not the determined period of the sound is within a predetermined range, and determining that the period of the sound is within the range, the detected sound is a detection target sound. And a determination unit for determining that there is.

  Further, in the sound detection method according to the present invention, the detection step of detecting a plurality of continuous sound signals from the input sound, and the inflection point of the waveform drawn by the trigonometric function match the detected sound. The period of the trigonometric function is varied so that the period of the waveform in which the inflection point matches the sound is determined as the period of the sound, and the determined period of the sound is within a predetermined range. Determining whether or not the plurality of detected physical sounds are detection target sounds when it is determined that the period of the physical sounds is within the range. Features.

  Further, the sound detection system according to the present invention is a sound detection system including a sound detection device and a monitoring center connected to the sound detection device via a network. A detection unit that detects a plurality of continuous sound signals, and a period of the trigonometric function is changed so that the inflection point of the waveform drawn by the trigonometric function matches the detected sound sounds. A period calculation unit that determines the period of the waveform that matches the inflection point as the period of the sound, and determines whether the determined period of the sound is within a predetermined range, and the period of the sound A determination unit that determines that the plurality of detected sound is a detection target sound when it is determined that the detected sound is within a range; and a case where the detected plurality of sound is determined to be the detection target sound Alert the monitoring center Characterized by comprising an output section for force, the.

  According to the present invention, there is an effect that the period of the detection target sound can be calculated even when the detection target sound is missing or the period fluctuates.

FIG. 1 is a block diagram illustrating a configuration of the sound detection device according to the first embodiment. FIG. 2 is a diagram illustrating an example of an acoustic signal and signal power. FIG. 3 is a diagram illustrating an example of a frequency structure of an acoustic signal. FIG. 4 is a diagram of an envelope obtained by an LPC cepstrum coefficient as an example of a feature amount of the frequency structure. FIG. 5 is a diagram showing probabilities that can be distinguished from footsteps in the same walk by the LPC cepstrum coefficient. FIG. 6 is a diagram illustrating an example of classification based on the frequency structure. FIG. 7 is an explanatory diagram of a general walking cycle. FIG. 8 is an explanatory diagram of the walking cycle in the present embodiment. FIG. 9 is an explanatory diagram illustrating a procedure for calculating a period of a sound by the period calculation unit. FIG. 10 is a diagram illustrating an example of a calculation result of a walking cycle by fitting a cosine function. FIG. 11 is a flowchart illustrating the procedure of the sound detection process performed by the sound detection device. FIG. 12 is a block diagram illustrating a configuration of the sound detection device according to the second embodiment. FIG. 13 is a flowchart illustrating the procedure of the sound detection process performed by the sound detection device.

  Exemplary embodiments of a sound detection device, a sound detection method, and a sound detection system according to the present invention will be described below in detail with reference to the accompanying drawings.

(First embodiment)
FIG. 1 is a block diagram illustrating a configuration of a sound detection device 100 according to the first embodiment. As shown in FIG. 1, a plurality of object sound detection devices 100 are arranged on a monitoring target and are connected to a microphone 200 that collects a sound generated from a sound source including the monitoring target. Although omitted in FIG. 1, the sound detection device 100 is connected to a monitoring center via a network. When the monitoring center receives an alarm from the sound detection device 100 indicating that an abnormality has been detected in the monitoring area, the monitoring center issues an instruction to the waiting guard to go to the monitoring area where the abnormality has been detected, It is a center that reports to related organizations such as the police and fire department as needed.

  The sound detection device 100 mainly includes an AD conversion unit 101, a sound detection unit 102, a classification unit 103, a period calculation unit 104, a determination unit 105, an alarm output unit 106, and a storage unit 110.

  The AD conversion unit 101 converts the sound collected from the microphone 200 into a digital sound and outputs it. Hereinafter, the sound collected by the microphone 200 is generically referred to as an analog sound and an acoustic signal including a digital sound after being converted by the AD converter 101, and when specifically indicated, the digital sound is referred to as acoustic data.

  The sound detection unit 102 detects continuous sound from the acoustic data converted by the AD conversion unit 101. Here, the physical sound is acoustic data in which signal intensity equal to or higher than a preset threshold is detected from the acoustic data output by the AD conversion unit 101, and includes footsteps. For example, the sound detection unit 102 calculates the signal power by obtaining the root mean square of the values of the acoustic data, and detects a signal power equal to or higher than a preset threshold as a sound. Note that the method for detecting a sound is not limited to this, and any method capable of detecting a sound from sound may be used.

  Here, a method for detecting footsteps from the same walking by the sound detection unit 102 will be described. The sound detection unit 102 obtains the power of the output signal with an analysis frame length of 50 ms and a frame interval of 2.5 ms, for example, and extracts the signal output time using 2.5 times the average power of background noise as a threshold value.

  FIG. 2 is a diagram illustrating an example of an acoustic signal and signal power. 2 shows the acoustic signal, the vertical axis shows the amplitude of the acoustic signal, and the horizontal axis shows the detection time of the acoustic signal. The lower part of FIG. 2 shows the signal power, the vertical axis shows the signal power intensity, and the horizontal axis shows the signal detection time.

  Also, the numbers on the horizontal axis indicate the number of output acoustic signals, and the horizontal lines in the figure indicate signal power thresholds for detecting sound from all the output acoustic signals. The sound detection unit 102 detects, as a sound, an acoustic signal output with a signal power equal to or higher than a threshold among the first to eighth acoustic signals output by the AD conversion unit 101. Here, the sound detection unit 102 detects a sound signal other than the fifth sound signal that does not satisfy the threshold as a sound.

  The classifying unit 103 classifies the sound detected by the sound detecting unit 102 into a group in which sound having a frequency structure similarity within a predetermined range. For example, the classification unit 103 compares the similarity of the frequency structure with the detected sound, and the sound having a high similarity is determined as the sound generated from the same sound source based on the frequency similarity information stored in the storage unit 110. Classify into one group.

  The classification unit 103 may obtain the frequency structure by Fourier transform or LPC cepstrum, and compare the similarities. FIG. 3 is a diagram illustrating an example of a frequency structure of an acoustic signal. The upper part of FIG. 3 is an example of an acoustic signal in which the vertical axis indicates amplitude and the horizontal axis indicates time. The lower part of FIG. 3 is an example of a frequency structure in which the vertical axis indicates signal power and the horizontal axis indicates frequency. For example, the classification unit 103 extracts a certain section in the acoustic signal shown in the upper part of FIG. 3, and calculates the frequency structure by Fourier transform. For the calculation of the frequency structure by the classification unit 103, other methods such as linear prediction analysis may be used in addition to the Fourier transform.

  FIG. 4 is a diagram of an envelope obtained by an LPC cepstrum coefficient as an example of a feature amount of the frequency structure. The classification unit 103 represents the calculated frequency structure with a polynomial, and calculates an envelope. The calculated envelope coefficient is used as the feature amount of the acoustic signal.

  The classification unit 103 also compares a plurality of feature amounts extracted from the frequency structure of the acoustic signal. For example, the classification unit 103 calculates the distance by obtaining the square difference of the feature amount, and calculates the probability that the same sound can be determined from the frequency similarity information stored in the storage unit 110 according to the distance. The probability that the sound can be identified as the same sound is a criterion indicating how much similarity is determined as the same sound. It should be noted that the predetermined range of similarity as a classification criterion can be arbitrarily set according to the installation environment of the sound detection device 100 such as the traffic volume of a person.

  Here, a method for obtaining the frequency similarity information stored in the storage unit 110 will be described. In general, since the frequency structures of footsteps in the same walk are similar, footsteps in the same walk can be classified by extracting a physical sound having a similar frequency structure. Therefore, for example, an LPC cepstrum coefficient is obtained as a frequency structure feature amount from footstep sample data prepared in advance, and the footstep distance distribution within the same walk and the footstep distance distribution between different walks are calculated. The probability that it can be discriminated from the footsteps of is calculated.

  FIG. 5 is a diagram showing probabilities that can be distinguished from footsteps in the same walk by the LPC cepstrum coefficient. In FIG. 5, the horizontal axis indicates the distance between the LPC cepstrum coefficients of footsteps, and the vertical axis indicates the probability that the footsteps are in the same walk. As an example, the classification unit 103 obtains the probability of footsteps in the same walk from the detected frequency structure distance of each sound by using this discrimination probability curve.

  The storage unit 110 stores physical sound data, frequency similarity data, cycle calculation reference data, footstep determination reference data, and the like. Here, the sound data is acoustic data classified by the classification unit 103. The frequency similarity data is data indicating characteristics of a predetermined frequency structure, and is information serving as a classification standard for collected sound.

  The cycle calculation reference data is data that is referred to when calculating the cycle. For example, there is a predetermined range of the cycle P that is changed when the cycle is calculated. Furthermore, the footstep determination reference data is data that is referred to when the detected sound is determined to be a footstep, and includes, for example, a predetermined range of the maximum value of the average value of the calculated cycles.

  FIG. 6 is a diagram illustrating an example of classification based on the frequency structure of acoustic signals. The signal power shown in FIG. 6 is the signal power shown in the lower part of FIG. 2, and group 1 and group 2 are groups of sound sounds classified by the classification unit 103.

  The classification unit 103 classifies the sound into group 1 and group 2 according to the similarity of the frequency structure of each detected signal power of the sound. Here, the classification unit 103 classifies the first acoustic signal, the third acoustic signal, and the seventh acoustic signal into the group 1 among the first to eighth acoustic signals based on the similarity comparison. ing. Similarly, the classification unit 103 classifies the second acoustic signal, the fourth acoustic signal, the sixth acoustic signal, and the eighth acoustic signal into group 2.

  The period calculation unit 104 calculates the period of the physical sound using the period calculation reference data stored in the storage unit 110 for each group classified by the classification unit 103. Here, the period of footsteps is calculated as the period of physical sounds. Here, the walking cycle will be described in calculating the cycle of footsteps. In general, the walking cycle is an operation from the first step of the first foot until the same foot contacts the ground again. FIG. 7 is an explanatory diagram of a general walking cycle. As shown in FIG. 7, for example, the grounding of the right foot that has been stepped first is grounded (1), the grounding of the left foot that has been stepped next is grounded (2), and the grounding of the right foot that has been stepped on again is grounded (3). In general, the walking cycle is one cycle from grounding (1) to grounding (3).

  Next, the walking cycle in the present embodiment will be described. The walking cycle in the present embodiment is an operation from when the first foot is stepped down until the next stepped foot touches down. FIG. 8 is an explanatory diagram of the walking cycle in the present embodiment. As in FIG. 7, for example, the grounding of the right foot that has been stepped first is grounded (1), the grounding of the left foot that has been stepped next is grounded (2), and the grounding of the right foot that has been stepped on again is grounded (3). In this case, the walking cycle in the present embodiment is one cycle from grounding (1) to grounding (2).

  The period calculation unit 104 varies the period value of the cosine function within a predetermined range, and applies the inflection point of the waveform drawn by the cosine function to the peak of the signal power of the sound that is classified into the same group. The period value of the cosine function is determined as the period of the sound.

ここで、ｙ_ｐ、ｎは、周期Ｐのとき、ｎ番目の物音がとる余弦関数の値とし、Ｐが示す周期として周期算出参照データが示す所定の範囲、例えば（０．１＜Ｐ＜２．０）が定められている。また、ｔ_ｎは最初に出力された物音を周期算出のための基準（時間ｔ_０＝０）とした場合における、ｎ番目に出力された物音の基準からの経過時刻（秒）とする。 For example, the period calculation unit 104 first determines the value of the cosine function by changing the period P of the following equation (1) within a predetermined range.

Here, yp _{, n} is a value of a cosine function taken by the nth sound in the period P, and a predetermined range indicated by the period calculation reference data as the period indicated by P, for example (0.1 <P <2 .0) is defined. Further, t _n is the elapsed time (seconds) from the reference of the nth output sound when the first output sound is used as a reference for period calculation (time t ₀ = 0).

Next, the period calculation unit 104 calculates an average value S _p of the value of the cosine function for the noises in the following (2) each period calculated by the equation.

Next, the period calculation unit 104 obtains the maximum value S _max of the average value S _p of the value of the cosine function calculated for each cycle. Period calculating section 104 is larger than a threshold maximum value S _max is predetermined, the period of time taking S _max and P _r, it is determined that if smaller periodicity no. Here, the time required for one step of walking is in the range of 0.3 to 0.9 seconds, and the value of S _max is set as the walking cycle when 0.3 <S _max <0.9 is satisfied.

  A method for calculating the period of a sound will be specifically described with reference to FIG. FIG. 9 is an explanatory diagram illustrating a procedure for calculating a period of a sound by the period calculation unit 104. The upper part of FIG. 9 shows the classified signal power. First, the period calculation unit 104 applies the time t when the peak of the signal power is output while changing the period P to the above equation (1).

For example, t of three sound sounds classified into the same group is t ₀ = 0 (second) (reference), t ₁ = 0.65 (second), and t ₂ = 1.9 (second), respectively. First, the period calculation unit 104 applies three points t to the equation (1) in which the period P = 0.2 is substituted, and obtains the value of the cosine function taken by each point.
y _{0.2, t0} = cos (2π × 0 / 0.2) = 1,
y _{0.2, t1} = cos (2π × 0.65 / 0.2) = 0,
_{y0.2, t2} = cos (2π × 1.9 / 0.2) = − 1.
The period calculation unit 104 obtains an average value from these values. here,
S _0.2 = (y _{0.2, t0} + y _{0.2, t1} + y _{0.2, t2} ) / 3 = 0.

Next, the period calculation unit 104 applies three points t to the equation (1) in which the period P = 0.6 is substituted, and obtains the value of the cosine function taken by each point.
y _{0.6, t0} = cos (2π × 0 / 0.6) = 1,
y _{0.6, t1} = cos (2π × 0.65 / 0.6) = 0.87,
_{y0.6, t2} = cos (2π × 1.9 / 0.6) = 0.5.
The period calculation unit 104 obtains an average value from these values. here,
_{S 0.6 = (y 0.6, t0} + y 0.6, t1 + y 0.6, t2) becomes /3=0.79.

  As shown in FIG. 9, the period calculation unit 104 obtains a period in which the value of the trigonometric function taken by each peak is maximum, and determines the obtained period as the period of a physical sound. Here, the average value when the period P = 0.2 is substituted is 0, and the average value when the period P = 0.6 is substituted is 0.79. 79. The period calculation unit 104 determines that the period P of 0.6 assigned to the group P when the maximum average value is 0.79 is the period of the sound of the group.

  When the period calculation unit 104 obtains the maximum period for one object sound period, the period of the next object sound is calculated by the series of processes as shown in FIG. Thereby, when a person's walk cycle changes, for example, when it changes from foot to run, a walk cycle is computable.

  FIG. 10 is a diagram illustrating an example of a calculation result of a walking cycle by fitting a cosine function. The lower part of FIG. 10 shows the signal power of the output footsteps, and the horizontal axis shows the elapsed time (seconds) from the output of the first signal power. The middle row indicates footsteps having a signal power equal to or higher than a threshold extracted from the output signal power. In the lower part of FIG. 10, in the vicinity of the elapsed time of 1.0 to 3.0 seconds, footsteps are generated twice in one step, and the data cannot be accurately calculated by the conventional cepstrum analysis. Further, in the vicinity of the elapsed time of 3.0 to 4.0 seconds, the output of the signal power less than the threshold is shown, and it is considered that the footstep sound collection is lost in this portion, and this is also detected by the conventional method. It is impossible data.

  The upper part of FIG. 10 shows a constant walking cycle calculated by trigonometric function fitting of the present embodiment to the extracted footstep data shown in the middle part. Here, the period calculation unit 104 indicates that the walking period is calculated as 0.71 seconds.

The determination unit 105 determines whether or not the period of the sound determined by the period calculation unit 104 is within a predetermined range indicated by the footstep determination reference data, and is classified when it is determined to be within the predetermined range. It is determined that the detected sound is the detection target sound. For example, if the period P _r is in the range of 0.3 to 0.9 seconds, which is predetermined as a predetermined range, determines classified sounds and human footstep is detected sound. Here, the detection target sound is a human footstep, but is not limited thereto, and may be a sound that destroys a wall or a door. Further, the predetermined range can be changed according to the situation.

  The alarm output unit 106 outputs an alarm when the sound classified by the determination unit 105 is determined to be a footstep. The alarm output unit 106 notifies the monitoring center that a footstep has been detected via the network.

  Next, the procedure of the sound detection process performed by the sound detection device 100 configured as described above will be described. FIG. 11 is a flowchart showing the procedure of the sound detection process performed by the sound detection device 100.

  The sound detection unit 102 detects a continuous sound from the digital sound converted by the AD conversion unit 101 (step S1). The sound detection unit 102 calculates the signal power by, for example, obtaining the root mean square of the value of the acoustic data, and determines whether or not the detected sound has a signal power equal to or higher than a preset threshold value ( Step S3).

  When it is determined that the sound detected by the sound detection unit 102 has a signal power equal to or higher than the threshold value (step S3: Yes), the classification unit 103 groups the sound according to the similarity of the frequency structure (step S4). . The classification unit 103 accumulates the sound data for each group in the storage unit 110 (step S5).

  The period calculation unit 104 determines whether or not data necessary for the period calculation has been accumulated (step S6). When it is determined that the number of data necessary for the period calculation is accumulated (step S6: Yes), the period calculation unit 104 changes the period P of the cosine function a predetermined number of times within a specified range (step S7). For example, 0.1 <P <2.0 is designated as the range of the period P.

  The period calculation unit 104 calculates a trigonometric function value y for each accumulated data (step S8). The period calculation unit 104 calculates the average value S of the trigonometric function values y calculated for each accumulated data (step S9). The cycle calculation unit 104 returns to step S7 and repeats the processing from step S7 to step S9 until the cycle P is changed a predetermined number of times. The predetermined number of times may be arbitrarily set according to the amount of data.

The cycle calculation unit 104 calculates the maximum value of the calculated average value S and the cycle _Pr at that time (step S10). Determination unit 105, _{P r} to determine whether it is within range of a threshold (step S11). Here, the determining unit 105, when P _r is in the range of a threshold, the detected sounds is determined to be footstep. Alarm output unit 106, if it is determined that _{P r} is in the range of a threshold by the determination unit 105 (step S11: Yes), outputs an alarm (step S12). Alternatively, the alarm output unit 106 notifies the monitoring center (step S13). The alarm output unit 106 may perform both alarm output and reporting to the monitoring center.

  As described above, according to the present embodiment, since the period calculation unit 104 calculates the period of a physical sound by fitting a trigonometric function, for example, even when a footstep that is a detection target sound is fluctuated or missing, The period of the detection target sound can be calculated.

(Second Embodiment)
In the first embodiment, acoustic signals in which signal intensity equal to or greater than a predetermined threshold is detected are classified into a predetermined group. In the present embodiment, the acoustic signals from which noise has been further removed are classified into a predetermined group among the acoustic signals in which the signal intensity equal to or greater than the predetermined threshold is detected.

  Specifically, depending on the installation environment of the present embodiment, periodic noise having periodicity such as excavation sound due to construction may occur. In the case of footsteps, it is difficult to create a feature model in advance because the frequency features differ depending on the type of footwear and flooring and other conditions. On the other hand, since periodic noise has a clear frequency characteristic, a model created in advance is used for filtering.

  FIG. 12 is a block diagram illustrating a configuration of the sound detection device 300 according to the second embodiment. As shown in FIG. 12, a plurality of object sound detection devices 300 are arranged on a monitoring target and connected to a microphone 200 that collects a sound generated from a sound source including the monitoring target. Although omitted in FIG. 12, the sound detection device 300 is connected to the monitoring center via the network in the same manner as the sound detection device 100. The microphone 200 is the same as that in the first embodiment.

  The sound detection device 300 includes an AD conversion unit 101, a sound detection unit 102, a noise removal unit 301, a classification unit 103, a period calculation unit 104, a determination unit 105, an alarm output unit 106, and a storage unit 110. The noise information storage unit 310 is mainly provided. Note that the functions and configurations of the units other than the noise removing unit 301 and the noise information storage unit 310 are the same as those in the first embodiment.

  The noise information storage unit 310 stores data indicating features of a predetermined frequency structure as noise frequency feature data. For example, a sound other than a footstep detected as a continuous specific sound includes a construction sound. In this case, the noise information storage unit 310 stores, as noise frequency feature data, a threshold value indicating the type and intensity of construction sound and a feature amount of noise. Note that the noise is not limited to this, and any noise other than the detection target may be used.

  The noise removal unit 301 removes noise from the sound that is detected as the specific sound by the sound detection unit 102. The noise removing unit 301 compares the frequency structure of the sound detected as the specific sound with the frequency characteristic data of the noise stored in the noise information storage unit 310, and matches the threshold value indicated by the noise frequency characteristic data. Sounds with noise are judged as noise and removed. For example, the noise removing unit 301 compares the feature amount of noise stored in the noise information storage unit 310 with the feature amount of the sound collected from the microphone 200, and is collected when the feature amount similarity is high. The sound is determined to be noise. This method is the same as the classification method performed by the classification unit 103 described in the first embodiment.

  As another example, since the footsteps that are detection targets of the present embodiment are generated from the floor surface of the monitoring range, the noise removing unit 301 collects only sound from the floor surface using a directional microphone or the like. By making a sound, it is possible to eliminate noise collection in other directions. A general gun microphone or a parabolic directional microphone can be applied to the microphone having directivity. In addition, a microphone array that can provide directivity by signal processing using a plurality of microphones may be applied.

  Next, the procedure of the sound detection process performed by the sound detection device 300 configured as described above will be described. FIG. 13 is a flowchart showing the procedure of the sound detection process performed by the sound detection device 300.

  The processing from step S21 to step S23 is the same as the processing from step S1 to step S3 in the flowchart of FIG. 11 described in the first embodiment.

  In step S24, the noise removal unit 301 determines whether or not the sound that has been determined by the sound detection unit 102 to be equal to or greater than the threshold value is not noise (step S24). For example, the noise removing unit 301 removes the sound from the classification target when the feature of the frequency structure of the sound that is determined to be equal to or higher than the threshold by the sound detecting unit 102 matches the frequency characteristic data of the noise.

  The processing from step S25 to step S34 is the same as the processing from step S4 to step S13 in the flowchart of FIG. 11 described in the first embodiment.

  In the present embodiment, footsteps have been described as detection target sounds, but the detection target sounds are not limited to this, and any sound having periodicity can be applied.

  As described above, according to the present embodiment, it is determined whether or not the characteristics of the noise frequency structure are the same as the frequency characteristic data of the noise among the collected sounds, and the sound determined to be matched is determined from the classification target. Since it is removed, false detection can be prevented.

DESCRIPTION OF SYMBOLS 100, 300 Object sound detection apparatus 101 AD conversion part 102 Object sound detection part 103 Classification part 104 Period calculation part 105 Judgment part 106 Alarm output part 110 Storage part 301 Noise removal part 310 Noise information storage part

Claims (8)

A detection unit that detects a plurality of continuous sound signals from the input sound;
The period of the trigonometric function is varied so that the inflection point of the waveform drawn by the trigonometric function matches the detected sound, and the period of the waveform at which the inflection point matches the sound A period calculation unit for determining the period;
It is determined whether or not the determined cycle of the sound is within a predetermined range, and when it is determined that the cycle of the sound is within the range, the plurality of detected sounds are detection target sounds. A determination unit for determining,
A sound detection device comprising: The period calculating unit substitutes the elapsed time from the detection time of the first object sound among the plurality of object sounds into the trigonometric function, and averages the values of the trigonometric functions calculated using the same object sound period. A value is obtained for each trigonometric function in which the value of the period is changed, and a maximum value is obtained from all the obtained average values,
The determination unit determines whether or not the acquired maximum value is within a predetermined range, and when the period of the sound is determined to be within the range, the detected sound is detected. Judging that it is the target sound,
The sound detection device according to claim 1. A classifying unit for classifying the detected sound objects based on frequency characteristics;
Further comprising
The period calculation unit determines a period of the sound for each of the plurality of classified sound;
The sound detection device according to claim 1. An output unit that outputs an alarm when it is determined that the plurality of detected sounds are the detection target sounds;
The sound detection device according to claim 1, further comprising: The detection target sound is a human footstep;
The sound detection device according to claim 3. It is determined whether or not the detected sound is noise based on the frequency characteristics of the detected sound, and when the detected sound is determined to be the noise, A noise removal unit for removing from the classification target by the classification unit;
The sound detection device according to claim 1, further comprising: A detection step of detecting a plurality of continuous sound signals from the input sound;
The period of the trigonometric function is varied so that the inflection point of the waveform drawn by the trigonometric function matches the detected sound, and the period of the waveform at which the inflection point matches the sound A period calculation step for determining the period;
It is determined whether or not the determined cycle of the sound is within a predetermined range, and when it is determined that the cycle of the sound is within the range, the plurality of detected sounds are detection target sounds. A determination step for determining
A sound detection method comprising: A noise detection system comprising a noise detection device and a monitoring center connected to the noise detection device via a network,
The sound detection device is
A detection unit that detects a plurality of continuous sound signals from the input sound;
The period of the trigonometric function is varied so that the inflection point of the waveform drawn by the trigonometric function matches the detected sound, and the period of the waveform at which the inflection point matches the sound A period calculation unit for determining the period;
It is determined whether or not the determined cycle of the sound is within a predetermined range, and when it is determined that the cycle of the sound is within the range, the plurality of detected sounds are detection target sounds. A determination unit for determining,
An output unit that outputs an alarm to the monitoring center when it is determined that the plurality of detected sounds are the detection target sounds;
A sound detection system characterized by comprising:

Images