CN108519149B - Tunnel accident monitoring and alarming system and method based on sound time-frequency domain analysis - Google Patents
Tunnel accident monitoring and alarming system and method based on sound time-frequency domain analysis Download PDFInfo
- Publication number
- CN108519149B CN108519149B CN201810267017.1A CN201810267017A CN108519149B CN 108519149 B CN108519149 B CN 108519149B CN 201810267017 A CN201810267017 A CN 201810267017A CN 108519149 B CN108519149 B CN 108519149B
- Authority
- CN
- China
- Prior art keywords
- sound
- tunnel
- module
- signals
- neural network
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01H—MEASUREMENT OF MECHANICAL VIBRATIONS OR ULTRASONIC, SONIC OR INFRASONIC WAVES
- G01H17/00—Measuring mechanical vibrations or ultrasonic, sonic or infrasonic waves, not provided for in the preceding groups
-
- G—PHYSICS
- G08—SIGNALLING
- G08B—SIGNALLING OR CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
- G08B25/00—Alarm systems in which the location of the alarm condition is signalled to a central station, e.g. fire or police telegraphic systems
- G08B25/01—Alarm systems in which the location of the alarm condition is signalled to a central station, e.g. fire or police telegraphic systems characterised by the transmission medium
- G08B25/10—Alarm systems in which the location of the alarm condition is signalled to a central station, e.g. fire or police telegraphic systems characterised by the transmission medium using wireless transmission systems
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Business, Economics & Management (AREA)
- Emergency Management (AREA)
- Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)
- Emergency Alarm Devices (AREA)
Abstract
The invention discloses a tunnel accident monitoring and alarming system and method based on sound time-frequency domain analysis, which comprises a sound acquisition and processing module, a DSP storage and analysis module and a control module, wherein the sound acquisition and processing module comprises a sound acquisition module, a DSP storage and analysis module: the sound acquisition processing module comprises a numerical converter and a sound sensor; the DSP storage analysis module comprises a ROM flash memory module, an SRAM data storage module and a DSP core processing module; according to the invention, the accident state in the tunnel is sensed through the sound signal, so that the tunnel operation mode is better adapted and the operation efficiency is improved, and the time-frequency domain analysis is carried out on the sound signal generated by the accident through the wavelet analysis and the improved neural network, so that the identification accuracy, the coverage range, the anti-interference degree and the signal-to-noise ratio of the accident signal are greatly improved; the invention can more comprehensively and directly acquire the tunnel accident information, achieve the overall monitoring of the tunnel, timely early warning, reduce the loss of people and property, meet the requirement of quick rescue and reduce the accident influence range.
Description
Technical Field
The invention belongs to the field of accident monitoring and wireless communication in tunnels, and relates to a tunnel accident monitoring and alarming system and method based on sound time-frequency domain analysis.
Background
In recent years, a large number of extra-long road tunnels are built and put into operation successively, the road tunnels are gradually shifted from a construction peak period to an operation peak period, the tunnels belong to national important infrastructure, and it is very important and necessary to maintain the safety of the tunnels, and therefore, the tunnels are strictly subjected to accident monitoring and management. However, the method is influenced by various aspects such as manual detection difficulty, technical operation and the like, so that the method becomes a first problem in the operation period and brings a difficult problem to tunnel operation management.
At present, the following problems exist in accident monitoring and alarming in a tunnel: a. the monitoring system has low network intelligence, most of the monitoring systems adopt fixed-point or manual inspection, and it is not practical to obtain real-time state information of a certain point in a tunnel; b. for a long tunnel, rescue arrangement and efficiency after an accident occurs are examined, and the tunnel is too long, so that the accident site is difficult to communicate with the outside, and emergency rescue activities are difficult to perform; c. the existing tunnel accident monitoring means mainly comprises smoke and video monitoring, but the coverage of a video monitoring distance and a video monitoring range is limited, the influence of light and internal traffic factors is large, the operation cost is high, the coverage of the smoke monitoring distance and the range is lower, the timeliness is poor, and the golden time of accident rescue is easily missed.
For example, the invention CN 104880245A provides a positioning alarm system based on vehicle impact noise, and the invention obtains
Calculating a characteristic value, wherein the algorithm has poor accuracy, low recognition degree of the impact signal and high false alarm rate; for example, CN 106887105 a and CN 103077609 a of the present invention respectively provide a tunnel monitoring system based on the characteristics of the disaster-stricken and multi-sensor sensing, which has high feasibility, cost, and construction technical requirements, and strong dependence on sensors and hardware facilities, and makes the actual construction environment and man-machine coordination ideal and separate from reality.
Disclosure of Invention
In order to overcome the problems in the prior art, the invention aims to provide a tunnel accident monitoring and alarming system and method based on sound time-frequency domain analysis.
In order to solve the problems in the prior art, the technical scheme of the invention is as follows:
the invention provides a tunnel accident monitoring and alarming method based on sound time-frequency domain analysis, which comprises the following steps:
step 1): collecting real-time sound signals in the tunnel, and screening effective sound signals as effective frames;
step 2): carrying out Fourier transformation and screening on the effective frame and the template library digital signal;
step 3): carrying out power spectrum conversion and screening on the effective frame and the template library digital signal;
step 4): performing wavelet decomposition on the effective frame and screening the effective frame to be used as a characteristic signal;
step 5): and (4) bringing the characteristic signals in the step (4) into a neural network for final judgment.
The method comprises the following specific steps:
step 1): screening the sound valid frame:
setting the frequency of the adopted sound signal to be 8000HZ, adopting a data frame which passes 50 points per frame and has a threshold value of 600 as an effective frame, and discarding a non-effective frame;
step 2): carrying out Fourier transform on the effective frame adopted in the step 1) and the characteristic signals in the template library and screening:
converting the sound characteristic signals correspondingly emitted when a tunnel collision accident occurs, automobile whistle sound and automobile engine sound characteristic signals into digital characteristic signals, storing the digital characteristic signals into a template base, carrying out Fourier transformation on the digital characteristic signals in an effective frame and the template base, converting the characteristic signals in a time domain into characteristic signals in a frequency domain, calculating correlation coefficients of the two functions after Fourier integral transformation, and keeping data with the correlation coefficients larger than a threshold value, otherwise, abandoning, and calculating the solution of the correlation coefficients according to the following formula:
in formula (1): cov (X, Y) represents the covariance formula, D (X) D (Y) represents the variance of X and Y, respectively;
step 3): performing power spectrum conversion and screening on the effective frame adopted in the step 1) and the characteristic signals in the template library:
processing discrete Fourier integral transformation with length of 1024 bytes, wherein the frequency is 8000Hz, converting digital characteristic signals in an effective frame and a template library into power spectrums, calculating correlation coefficients of the converted functions of the two power spectrums, further calculating effective frame signals with the correlation coefficients exceeding a threshold value, and calculating the power spectrums according to the following formula:
in formula (2): s (ω) represents the power spectrum of the active frame, x (t) represents the time domain signal, and P represents the power spectral density;
step 4): wavelet decomposition of effective frame signals, retention of characteristic signals:
correspondingly converting all the effective frames exceeding the threshold in the step 3) into a fixed reasonable interval, namely normalizing, performing wavelet decomposition on effective frame signals exceeding the threshold, decomposing numerical signals of different time domains and frequency domains, removing high-frequency signals, reserving decomposed low-frequency signals, gradually decomposing the reserved low-frequency signals through a Mallat algorithm, and taking the reserved low-frequency signals as characteristic signals, wherein a wavelet decomposition formula is expressed as the following formula:
in formula (4): h represents the filter coefficient, CjnA scale coefficient representing a length space;
step 5): training a three-layer neural network by using data for feature extraction based on wavelet decomposition, and putting the feature signals retained in the step 4) into the neural network for final judgment; specifically training the neural network process: training the automobile whistle and the automobile engine sound together with the tunnel impact sound, wherein the tunnel impact sound, the automobile whistle and the automobile engine sound are decomposed through wavelets and then serve as training samples for standby; seventy-five percent of samples are extracted in a cross validation mode to serve as training data, and the rest are testing data; adopting three layers of neural networks, wherein an input layer is a sample characteristic value, the number of hidden layers is greater than that of the input layers, and an output layer is an identification result; initializing the weight value and the threshold value theta (W, b) of the neural network, putting a training sample into the neural network for iterative computation, adjusting the weight value W and the threshold value b, bringing test data into the neural network for classification in the trained neural network, and taking the weight value and the threshold value of the neural network with the best performance as a final neural network classifier;
step 6): inputting a real-time characteristic signal to judge an output state result by a neural network:
and (3) running the real-time sound signals collected by the sound sensor 2 in the tunnel according to the sequence of the steps 1 to 4, substituting the characteristic signals obtained after running into the neural network model constructed in the step 5 for calculation, and judging the real-time tunnel state classification.
In step 5, the weight W and the threshold b are finely adjusted according to the following formula:
in formula (10): wijIndicating the corresponding weights, α indicating the convergence rate,
representing error function to calculate partial derivative of weight;
in formula (11):
which is indicative of a corresponding threshold value, is,
representing the partial derivative of the error function to the threshold value;
in step 5, the weight W and the threshold b are finely adjusted according to the following formula:
in formula (13): x is the number ofiThe number of neurons in this layer of the neural network is represented.
In the three-layer neural network, an input layer is 11 neurons, a hidden layer is 15 neurons, and an output layer is 3 neurons.
The invention also provides a tunnel accident monitoring and alarming system based on sound time-frequency domain analysis, which comprises a sound acquisition and processing module, a DSP storage and analysis module and a control module, wherein:
the sound acquisition processing module comprises a numerical converter and a sound sensor;
the DSP storage analysis module comprises a ROM flash memory module, an SRAM data storage module and a DSP core processing module;
the control module comprises an alarm module, a communication module and a positioning module, wherein the communication module comprises a communication control device and a communication transmission device, the positioning module is used for determining the position information of the accident vehicle, and the alarm module comprises an alarm signal device, an alarm control system and an alarm communication device;
the numerical converter is connected with the sound sensor and the DSP storage and analysis module; the DSP storage and analysis module is connected with the numerical converter and the control module; the control module is connected with the DSP core processing module and the tunnel monitoring center, the communication module is connected with a communication system of the tunnel monitoring center, the alarm module is connected with the emergency fire alarm system through a relay interface, and the positioning module is connected with the sound sensor and is arranged in the sound sensor;
the sound sensor is used for collecting sound in the tunnel, and the collected sound is converted into a digital signal through the numerical value converter and then is transmitted to the SRAM data storage module; the DSP core processing module is used for loading codes in the ROM flash memory, executing the codes, reading data in the SRAM data storage module and sending an instruction obtained after operation to the communication transmission device.
DSP stores analysis module and control module and all settles in tunnel surveillance center, and in the sound collection processing module, numerical value converter settled in tunnel surveillance center, sound sensor arranged in the tunnel inner wall.
The sound sensors are arranged on the side walls on the two sides of the tunnel, the height of the side walls is 2-2.5 m, and multiple groups of sound sensors are set at intervals of preset distances along the extending direction of the sound sensors.
The sound sensor adopts an ARM9 sound sensor, the numerical converter adopts an ads5422 conversion chip, the DSP core processing module adopts a TMS320C54 DSP board, and the ROM flash memory module adopts SST39LF/VF160, and is a 1M16bit CMOS multifunctional FlashMPF device.
Compared with the prior art, the invention has the following advantages: the invention senses the accident state in the tunnel through the sound signal, better adapts to the tunnel operation mode and improves the operation efficiency, and also provides a new mode for monitoring and alarming the tunnel accident; according to the invention, the wavelet analysis and the improved neural network are used for carrying out time-frequency domain analysis on the sound signals generated by the accident, so that the identification accuracy, the coverage area, the anti-interference degree and the signal-to-noise ratio of the accident signals are greatly improved; the invention can more comprehensively and directly acquire the tunnel accident information, thereby achieving the overall monitoring of the tunnel, timely early warning, reducing the loss of people and property, satisfying the requirement of quick rescue and reducing the accident influence range; training a three-layer neural network by using data extracted by features based on wavelet decomposition, putting feature signals into the neural network for final judgment, training automobile whistle and automobile engine sound together with tunnel impact sound, taking the tunnel impact sound, the automobile whistle and the automobile engine sound as training samples for standby after the wavelet decomposition, wherein the wavelet analysis is a time (space) frequency localization analysis, the time domain analysis is a method for directly analyzing a system in a time domain according to a time domain expression of output quantity, and the frequency domain analysis is a method for decomposing a time history waveform into a plurality of single harmonic components through Fourier transform so as to obtain a frequency structure of the signals and information of each harmonic and phase, so that the accuracy and the recognition speed of an algorithm are improved;
in addition, according to the invention, through reasonable formula design, a risk coefficient Rc is added on the right sides of the formulas (10) and (11), so that the probability of neural network overfitting is reduced, and the risk of easy damage is avoided while signals are effectively collected due to reasonable position arrangement of the sound sensors.
Drawings
FIG. 1 is a schematic diagram of the system of the present invention;
FIG. 2 is a schematic diagram of the connection of accident monitoring alarm hardware according to an embodiment of the present invention;
fig. 3 is a flow chart of an improved algorithm according to an embodiment of the present invention.
In the figure, 1-tunnel, 2-sound sensor, 3-numerical converter, 4-ROM flash module, 5-SRAM data storage module, 6-alarm signal device, 7-alarm control system, 8-alarm communication device, 9-communication control device, 10-communication transmission device and 11-DSP core processing module.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
Referring to fig. 1 and 2, the tunnel accident alarm system based on sound time-frequency domain analysis of the present invention includes a sound collection processing module, a DSP storage analysis module, and a control module; the sound acquisition processing module comprises a sound sensor 2 and a numerical converter 3, the sound sensor 2 is used for collecting sound signals in the tunnel 1, and the numerical converter 3 amplifies the electric signals through digital-to-analog conversion and converts the electric signals into digital signals; the DSP storage and analysis module comprises a ROM flash memory module 4, an SRAM data storage module 5 and a DSP core processing module 11, the SRAM data storage module 5 stores all digital signals and waits for the DSP core processing module 11 to process, the ROM flash memory module 4 stores an identification library and an executable file, and the DSP core processing module 11 performs identification comparison and analysis through the SRAM data storage module 5 and the file stored in the ROM flash memory module 4; the control module comprises an alarm module, a communication module and a positioning module, wherein the communication module comprises a communication control device 9 and a communication transmission device 10 and is used for coordinating accident information received by a management and control center and monitoring alarm, the positioning module is used for determining position information of an accident vehicle, and the alarm module comprises an alarm signal device 6, an alarm control system 7 and an alarm communication device 8 and is used for sending accident warning information to inform the monitoring center of linkage with a disaster prevention rescue department.
Furthermore, the sound sensors 2 are distributed on the whole road section or the important road section of the tunnel 1, are arranged on the side walls at two sides of the tunnel 1 and are at positions approximately equal to the height of the automobile, and are set into a plurality of groups at preset intervals along the extending direction of the sound sensors, the monitoring ranges of the adjacent sound sensors 2 are overlapped by referring to the folding type covering, so that barrier-free and blind-area-free collection is realized, the on-site adjustment can also be carried out according to the sensing property of the sound sensors 2, and the sound sensors 2 adopt ARM9 sound sensors.
Further, the numerical value converter 3 is arranged in a tunnel monitoring center, is connected with the sound sensor 2 and the DSP storage and analysis module, converts the electric signals received from the sound sensor 2 into data signals and transmits the data signals to the DSP storage and analysis module, and the numerical value converter 3 adopts an ads5422 conversion chip.
Furthermore, the DSP storage analysis module is connected with the numerical converter 3 and the control module, is arranged in the tunnel monitoring center, receives the digital signals transmitted by the numerical converter 3, performs comparative analysis, and sends the processing results to the corresponding control module; the DSP core processing module 11 adopts a TMS320C54 DSP board, and the ROM flash memory module 4 adopts SST39LF/VF160, which is a 1M16bit CMOS multifunctional FlashMPF device.
Further, the control module is arranged in a tunnel monitoring center, is connected with the DSP core processing module 11 and related traffic managers and rescue departments, and carries out emergency measures through transmitted accident state information. The communication module is connected with a communication system of a tunnel monitoring center, the alarm module is connected with a related emergency fire alarm system through a relay interface, the positioning module is connected with the sound sensor 2 and arranged inside the sound sensor 2, the sound sensor 2 corresponding to an accident site is identified through audio digital information, then the accident site information is determined through the positioning module, and the positioning module adopts Mtk or an Mstar GPS chip.
Further, when the DSP core processing module 11 receives the sound signal transmitted by the sound collection module, the sound signal is first subjected to noise cancellation processing by an algorithm through preliminary analysis of time domain characteristics, so that the recognition rate is more than 60%; then, through wavelet analysis and feature extraction of the improved neural network in a time domain and a frequency domain, the recognition degree of the collision signal is continuously improved on the basis of being recognized by the time domain, so that the recognition rate of the collision signal is up to more than 90%; when the noise signal is higher than the threshold value, the DSP core processing module 11 is linked with the control module immediately to alarm and rescue.
The DSP core processing module 11 is the core of the whole tunnel accident monitoring and alarming system, and completes the functions of collecting, controlling, storing, processing and communicating with the outside, and the like of the audio signal, and the invention mainly lies in the improvement of the DSP core processing module 11 aiming at the sound signal processing algorithm:
step 1): the frequency of the sound signals adopted in the improved algorithm is set to be 8000HZ, and because the quantity of the sound signals collected in the tunnel is huge, in order to reduce the calculated quantity and improve the accuracy, the improved algorithm screens the sound signals in a time domain, and adopts a data frame which passes 50 points per frame and has a threshold value of 600 as an effective frame, and discards an ineffective frame.
Step 2): converting the sound characteristic signals correspondingly emitted when a tunnel collision accident occurs, automobile whistle sound characteristic signals and automobile engine sound characteristic signals into digital characteristic signals, storing the digital characteristic signals into a template base, performing Fourier transformation on the effective frames and the digital characteristic signals in the template base, converting the characteristic signals in a time domain into characteristic signals in a frequency domain, calculating correlation coefficients of functions obtained after the Fourier integral transformation, and retaining data with the correlation coefficients larger than a threshold value, otherwise, discarding the data. The solution to the correlation coefficient is calculated according to:
in formula (1): cov (X, Y) represents the covariance formula, D (X) D (Y) represents the variance of X and Y, respectively.
Step 3): processing discrete Fourier integral transformation with length of 1024 bytes, wherein the frequency is 8000Hz, converting digital characteristic signals in an effective frame and a template library into power spectrums, calculating correlation coefficients of the converted functions of the two power spectrums, further calculating effective frame signals with the cross correlation coefficients exceeding a threshold value, and calculating the power spectrums according to the following formula:
in formula (2): s (ω) represents the power spectrum of the active frame, x (t) represents the time domain signal, and P represents the power spectral density.
Step 4): all effective frames exceeding the threshold are correspondingly converted into a fixed reasonable interval, namely normalization, wavelet decomposition is carried out on effective frame signals exceeding the threshold, numerical value signals in different time domains and frequency domains are decomposed, high-frequency signals are removed, decomposed low-frequency signals are reserved, the reserved low-frequency signals are gradually decomposed through a Mallat algorithm and serve as characteristic signals, and a wavelet decomposition formula is represented as the following formula:
in formula (4): h represents the filter coefficient, CjnScale coefficients representing a length space.
Step 5): and (4) carrying out wavelet decomposition on the tunnel impact, the automobile whistle and the automobile engine sound characteristic signals in the template library according to the step (4), taking the digital characteristic signals decomposed into different frequency spaces as samples, taking 75% of the samples as training samples to construct a neural network, and taking the rest 25% of the samples as test samples to detect errors of the neural network. The improved algorithm adopts three layers of neural networks, wherein an input layer is a sample characteristic value, the number of hidden layers is greater than that of the input layers, and an output layer is a recognition result of W and b. Initializing the weight value of the neural network and a threshold value theta (W, b), putting the training sample into the neural network for iterative computation, and adjusting the weight value W and the threshold value b as represented by the following formula:
z(2)＝W(1)x+b(1) (5)
a(2)＝f(z(2)) (6)
z(3)＝W(2)a(2)+b(2) (7)
D＝f(z(3)) (8)
in formulae (5) and (7): w (1) and W (2) represent weights, and b (1) and b (2) represent thresholds.
In formulae (6) and (8): a (2) represents the value obtained by calculating the threshold value of the neural network by the training sample, and D represents the stage value of one neural network iteration.
Further, the improved algorithm is based on the neural network analysis of the traditional back propagation error, and the weight W and the threshold b are finely adjusted, and the formula is as follows:
in formula (10): wijIndicating the corresponding weights, α indicating the convergence rate,
and representing the deviation of the error function to the weight.
In formula (11):
which is indicative of a corresponding threshold value, is,
indicating that the error function makes a partial derivative of the threshold.
By adopting a Widrow-Hoff learning rule, an error function formula is expressed as follows:
in formula (12): djRepresenting the output value, y, of one iterationiThe initial true value is represented, and the adjustment range of the weight W and the threshold b is analyzed through an error function by the improved algorithm.
Further, the improved algorithm adds a risk coefficient Rc to the right side of equations (10) and (11) to reduce the probability of overfitting the neural network, where Rc is expressed as follows:
in formula (13): x is the number ofiThe number of features is shown.
And performing wavelet decomposition on the tunnel impact, the automobile whistle and the automobile engine sound characteristic signals in the template library, and optimizing the weight W and the threshold b to construct a neural network capable of judging the sound signals and performing state classification.
Step 6): and (3) operating the real-time sound signals collected by the sound sensor 2 in the tunnel according to the steps 1-4, substituting the real-time sound signals into the neural network model constructed in the step 5 for calculation, and judging real-time tunnel state classification.
Referring to fig. 3, the algorithm flow chart describes a method for recognizing whether a collision accident occurs in a tunnel through a sound signal by using the tunnel accident monitoring and warning system, and the steps are as follows:
1. collecting real-time sound signals in the tunnel;
2. screening out effective sound signals as effective frames;
3. carrying out Fourier transformation and screening on the effective frame and the template library digital signal;
4. carrying out power spectrum conversion and screening on the effective frame and the template library digital signal;
5. performing wavelet decomposition on the effective frame and screening the effective frame to be used as a characteristic signal;
6. and judging the output state result by the neural network through the characteristic signal.
In the whole DSP core processing module processing process, the improved algorithm can greatly improve the recognition efficiency of the accident sound signals and improve the recognition accuracy, the anti-interference degree and the signal-to-noise ratio of the accident sound signals by analyzing and recognizing a plurality of domains and characteristics of digital signals, analyzing time and frequency domains by wavelet analysis and the improved neural network, and extracting by Matlab simulation.
The above description is further intended to illustrate the process of the present invention in detail with reference to specific examples, which should not be construed as limiting the practice of the process of the present invention. For a person skilled in the art to which the invention pertains, several equivalent alternatives or obvious modifications, all of which have the same properties or uses, without departing from the inventive concept, should be considered as falling within the scope of the patent protection of the invention, as determined by the claims filed.
Claims (8)
1. A tunnel accident monitoring and alarming method based on sound time-frequency domain analysis is characterized by comprising the following steps:
step 1): collecting real-time sound signals in the tunnel, and screening effective sound signals as effective frames;
step 2): carrying out Fourier transformation and screening on the effective frame and the template library digital signal;
step 3): carrying out power spectrum conversion and screening on the effective frame and the template library digital signal;
step 4): performing wavelet decomposition on the effective frame and screening the effective frame to be used as a characteristic signal;
step 5): bringing the characteristic signals in the step 4) into a neural network for final judgment;
the method comprises the following specific steps:
step 1): screening the sound valid frame:
setting the frequency of the adopted sound signal to be 8000HZ, adopting a data frame which passes 50 points per frame and has a threshold value of 600 as an effective frame, and discarding a non-effective frame;
step 2): carrying out Fourier transform and screening on the effective frame and the template library digital signal adopted in the step 1):
converting a sound characteristic signal correspondingly emitted when a tunnel collision accident occurs, an automobile whistle sound and an automobile engine sound signal into digital signals, storing the digital signals into a template library, carrying out Fourier transformation on an effective frame and the template library digital signals, converting the characteristic signals on a time domain into the characteristic signals on a frequency domain, calculating a correlation coefficient of a function obtained by the two Fourier transformations, reserving data with the correlation coefficient larger than a threshold value, otherwise, abandoning the data, and calculating the solution of the correlation coefficient according to the following formula:
in formula (1): cov (X, Y) represents the covariance formula, D (X), D (Y) represent the variances of X and Y, respectively;
step 3): performing power spectrum conversion and screening on the effective frame and the template library digital signal adopted in the step 1):
processing discrete Fourier integral transformation with length of 1024 bytes, wherein the frequency is 8000Hz, converting effective frame and template library digital signals into power spectrums, calculating correlation coefficients of functions obtained after the two power spectrums are converted, further calculating effective frame signals with the correlation coefficients exceeding a threshold value, and calculating the power spectrums according to the following formula:
in formula (2): s (ω) represents the power spectrum of the active frame, x (t) represents the time domain signal, and P represents the power spectral density;
step 4): wavelet decomposition of effective frame signals, retention of characteristic signals:
correspondingly converting all the effective frames exceeding the threshold in the step 3) into a fixed reasonable interval, namely normalizing, performing wavelet decomposition on effective frame signals exceeding the threshold, decomposing numerical signals of different time domains and frequency domains, removing high-frequency signals, reserving decomposed low-frequency signals, gradually decomposing the reserved low-frequency signals through a Mallat algorithm, and taking the reserved low-frequency signals as characteristic signals, wherein a wavelet decomposition formula is expressed as the following formula:
in formula (4): h represents the filter coefficient, CjnA scale coefficient representing a length space;
step 5): training a three-layer neural network by using data for feature extraction based on wavelet decomposition, and putting the feature signals retained in the step 4) into the neural network for final judgment; specifically training the neural network process: training the automobile whistle and the automobile engine sound together with the tunnel impact sound, wherein the tunnel impact sound, the automobile whistle and the automobile engine sound are decomposed through wavelets and then serve as training samples for standby; seventy-five percent of samples are extracted in a cross validation mode to serve as training data, and the rest are testing data; adopting three layers of neural networks, wherein an input layer is a sample characteristic value, the number of hidden layers is greater than that of the input layers, and an output layer is an identification result; initializing the weight value and the threshold value theta (W, b) of the neural network, putting a training sample into the neural network for iterative computation, adjusting the weight value W and the threshold value b, bringing test data into the neural network for classification in the trained neural network, and taking the weight value and the threshold value of the neural network with the best performance as a final neural network classifier;
step 6): inputting a real-time characteristic signal to judge an output state result by a neural network:
and (3) sequentially operating the real-time sound signals collected by the sound sensor (2) in the tunnel according to the steps 1) to 4), substituting the characteristic signals obtained after operation into the neural network model constructed in the step 5) for calculation, and judging the real-time tunnel state classification.
2. The method for monitoring and alarming tunnel accidents based on sound time-frequency domain analysis according to claim 1, wherein in the step 5), the weight W and the threshold b are finely adjusted according to the following formula:
in formula (10): wijIndicating the corresponding weights, α indicating the convergence rate,
representing error function to calculate partial derivative of weight;
in formula (11):
which is indicative of a corresponding threshold value, is,
indicating that the error function partially derives from the threshold.
3. The method for monitoring and alarming tunnel accidents based on sound time-frequency domain analysis according to claim 1, wherein in the step 5), the weight W and the threshold b are finely adjusted according to the following formula:
in formula (13): x is the number ofiTo representThe number of neurons in this layer of the neural network.
4. The method according to claim 1, wherein in the three-layer neural network, an input layer comprises 11 neurons, a hidden layer comprises 15 neurons, and an output layer comprises 3 neurons.
5. The system for monitoring and alarming tunnel accidents based on sound time-frequency domain analysis according to any one of claims 1 to 4, is characterized by comprising a sound collection processing module, a DSP storage analysis module and a control module, wherein:
the sound acquisition processing module comprises a numerical converter (3) and a sound sensor (2);
the DSP storage analysis module comprises a ROM flash memory module (4), an SRAM data storage module (5) and a DSP core processing module (11);
the control module comprises an alarm module, a communication module and a positioning module, wherein the communication module comprises a communication control device (9) and a communication transmission device (10), the positioning module is used for determining the position information of the accident vehicle, and the alarm module comprises an alarm signal device (6), an alarm control system (7) and an alarm communication device (8);
the numerical value converter (3) is connected with the sound sensor (2) and the DSP storage and analysis module; the DSP storage and analysis module is connected with the numerical value converter (3) and the control module; the control module is connected with the DSP core processing module (11) and the tunnel monitoring center, the communication module is connected with a communication system of the tunnel monitoring center, the alarm module is connected with the emergency fire alarm system through a relay interface, and the positioning module is connected with the sound sensor (2) and is arranged in the sound sensor;
the sound sensor (2) is used for collecting sound in the tunnel, and the collected sound is converted into a digital signal through the numerical value converter (3) and then is transmitted to the SRAM data storage module (5); the DSP core processing module is used for loading codes in the ROM flash memory, executing the codes, reading data in the SRAM data storage module (5) and sending an instruction obtained after operation to the communication transmission device (10).
6. The system for monitoring and alarming tunnel accidents based on sound time-frequency domain analysis according to claim 5, wherein the DSP storage and analysis module and the control module are both arranged in a tunnel monitoring center, the sound collection and processing module, the numerical converter (3) is arranged in the tunnel monitoring center, and the sound sensor (2) is arranged on the inner wall of the tunnel.
7. The system for monitoring and alarming tunnel accidents based on sound time-frequency domain analysis according to claim 6, wherein the sound sensors (2) are arranged at the positions of 2-2.5 m of the height of the side walls at two sides of the tunnel and are set into a plurality of groups at preset intervals along the extending direction of the sound sensors.
8. The system for monitoring and alarming the tunnel accident based on sound time-frequency domain analysis is characterized in that a sound sensor (2) adopts an ARM9 sound sensor, a numerical converter (3) adopts an ads5422 conversion chip, a DSP core processing module (11) adopts a TMS320C54 DSP board, and a ROM flash memory module (4) adopts SST39LF/VF160 and is a 1M16bit CMOS multifunctional FlashMPF device.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810267017.1A CN108519149B (en) | 2018-03-28 | 2018-03-28 | Tunnel accident monitoring and alarming system and method based on sound time-frequency domain analysis |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810267017.1A CN108519149B (en) | 2018-03-28 | 2018-03-28 | Tunnel accident monitoring and alarming system and method based on sound time-frequency domain analysis |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108519149A CN108519149A (en) | 2018-09-11 |
| CN108519149B true CN108519149B (en) | 2020-09-01 |
Family
ID=63430738
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810267017.1A Active CN108519149B (en) | 2018-03-28 | 2018-03-28 | Tunnel accident monitoring and alarming system and method based on sound time-frequency domain analysis |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108519149B (en) |
Families Citing this family (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109695475A (en) * | 2018-11-30 | 2019-04-30 | 太原理工大学 | A kind of coal mine roadway exception monitoring device and method based on sound |
| CN109616140B (en) * | 2018-12-12 | 2022-08-30 | 浩云科技股份有限公司 | Abnormal sound analysis system |
| CN109754581A (en) * | 2019-01-15 | 2019-05-14 | 深圳市宇航智造技术有限公司 | Calculator room equipment method for early warning, device, equipment and storage medium |
| CN109597342B (en) * | 2019-01-16 | 2020-10-20 | 郑州轻工业学院 | Dynamic networking intelligent identification sand production ship monitoring device and method |
| CN111862519A (en) * | 2020-08-17 | 2020-10-30 | 浙江中辰安全设备有限公司 | Fire early warning fog calculation service gateway |
| CN112951271A (en) * | 2020-11-27 | 2021-06-11 | 长沙理工大学 | Tunnel traffic safety warning method and system based on acoustic assistance |
| CN112881744B (en) * | 2021-02-08 | 2023-04-07 | 中科(湖南)先进轨道交通研究院有限公司 | Train speed and direction measuring device based on train sound and working method thereof |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102741896A (en) * | 2009-12-21 | 2012-10-17 | 电话有限公司 | Portable apparatus and method for detecting and notifying of vehicle accidents |
| CN105913663A (en) * | 2016-06-30 | 2016-08-31 | 南京嘉瑞斯电力科技有限公司 | Automobile accident noise detection and positioning system based on sound sensor |
| CN105989748A (en) * | 2015-03-20 | 2016-10-05 | 现代自动车株式会社 | Accident information management appratus, vehicle including the same, and accident information management method |
Family Cites Families (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2002008189A (en) * | 2000-06-22 | 2002-01-11 | Matsushita Electric Ind Co Ltd | Vehicle detector and vehicle detection method |
| CN101710450A (en) * | 2009-12-21 | 2010-05-19 | 金会庆 | Traffic accident automatic sound detection and alarm terminal |
| CN101877172B (en) * | 2009-12-21 | 2011-12-14 | 金会庆 | Impact sound-based automatic traffic accident detection method |
| CN102163427B (en) * | 2010-12-20 | 2012-09-12 | 北京邮电大学 | Method for detecting audio exceptional event based on environmental model |
| CN102568153A (en) * | 2010-12-20 | 2012-07-11 | 淄博高新区联创科技服务中心 | Vehicle collision detection device |
| CN102122133B (en) * | 2011-01-21 | 2012-10-10 | 北京工业大学 | Self-adaption wavelet neural network abnormity detection and fault diagnosis classification system and method |
| CN202362932U (en) * | 2011-11-28 | 2012-08-01 | 长安大学 | Information monitoring device of traffic accident scene |
| CN103622707A (en) * | 2012-08-20 | 2014-03-12 | 王震 | Traffic safety detecting system and method |
| CN103077609A (en) * | 2012-12-26 | 2013-05-01 | 上海市城市建设设计研究总院 | Tunnel traffic accident monitoring method and system based on sensing of multiple sensors |
| CN104102923A (en) * | 2014-07-16 | 2014-10-15 | 西安建筑科技大学 | Nipponia nippon individual recognition method based on MFCC algorithm |
| CN106841403A (en) * | 2017-01-23 | 2017-06-13 | 天津大学 | A kind of acoustics glass defect detection method based on neutral net |
| CN107086036A (en) * | 2017-04-19 | 2017-08-22 | 杭州派尼澳电子科技有限公司 | A kind of freeway tunnel method for safety monitoring |
| CN207008862U (en) * | 2017-07-11 | 2018-02-13 | 杭州天铂红外光电技术有限公司 | Detect the device of tunnel accident |
| CN107545890A (en) * | 2017-08-31 | 2018-01-05 | 桂林电子科技大学 | A kind of sound event recognition method |
| CN107705633B (en) * | 2017-11-06 | 2020-07-28 | 张定宇 | System and method for preventing secondary accidents of automobiles in expressway tunnel |
-
2018
- 2018-03-28 CN CN201810267017.1A patent/CN108519149B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102741896A (en) * | 2009-12-21 | 2012-10-17 | 电话有限公司 | Portable apparatus and method for detecting and notifying of vehicle accidents |
| CN105989748A (en) * | 2015-03-20 | 2016-10-05 | 现代自动车株式会社 | Accident information management appratus, vehicle including the same, and accident information management method |
| CN105913663A (en) * | 2016-06-30 | 2016-08-31 | 南京嘉瑞斯电力科技有限公司 | Automobile accident noise detection and positioning system based on sound sensor |
Non-Patent Citations (1)
| Title |
|---|
| 基于音频检测的交通事故自动报警系统研究;凤鹏飞等;《唐山学院学报》;20170530;第30卷(第3期);第50-54页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108519149A (en) | 2018-09-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108519149B (en) | Tunnel accident monitoring and alarming system and method based on sound time-frequency domain analysis | |
| CN102045120B (en) | Vibration signal identification method for optical fiber perimeter system | |
| CN103033567B (en) | Pipeline defect signal identification method based on guided wave | |
| CN108133295B (en) | Motor vehicle exhaust concentration continuous time prediction method for target road section | |
| CN108445868B (en) | Intelligent automobile fault diagnosis system and method based on modern signal processing technology | |
| CN108197731B (en) | Motor vehicle exhaust remote measurement and vehicle inspection result consistency method based on co-training | |
| CN110570613A (en) | Fence vibration intrusion positioning and mode identification method based on distributed optical fiber system | |
| US20090115635A1 (en) | Detection and classification of running vehicles based on acoustic signatures | |
| CN102682765A (en) | Expressway audio vehicle detection device and method thereof | |
| CN109816987A (en) | A kind of automobile whistle electronic police enforces the law capturing system and its grasp shoot method | |
| CN112541480B (en) | Online identification method and system for tunnel foreign matter invasion event | |
| CN110188737A (en) | The thermal runaway method for early warning of acoustical signal processing is opened based on lithium cell safety valve | |
| CN108848571B (en) | Rail transit safety monitoring system and monitoring method based on MEMS sensor | |
| CN116517860A (en) | Ventilator fault early warning system based on data analysis | |
| CN106382981A (en) | Single station infrasonic wave signal recognition and extraction method | |
| CN111341334A (en) | Noise reduction and abnormal sound detection system and method applied to rail transit | |
| CN108506736A (en) | A kind of invasion signal diagnostic system and method | |
| CN105352541B (en) | A kind of transformer station high-voltage side bus auxiliary monitoring system and its monitoring method based on power network disaster prevention disaster reduction system | |
| Smailov et al. | A novel deep CNN-RNN approach for real-time impulsive sound detection to detect dangerous events | |
| CN112382051B (en) | Wisdom house security protection system based on block chain | |
| CN111951505B (en) | Fence vibration intrusion positioning and mode identification method based on distributed optical fiber system | |
| CN206781702U (en) | A kind of speech recognition automotive theft proof system based on quantum nerve network | |
| CN112948715A (en) | Vehicle classification method based on short-time GPS track data | |
| CN117452496A (en) | Target detection and identification method for seismoacoustic sensor | |
| CN104374885A (en) | Explosive gas dangerous sound resource function detection device |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |