CN107545890A - A kind of sound event recognition method - Google Patents
A kind of sound event recognition method Download PDFInfo
- Publication number
- CN107545890A CN107545890A CN201710776733.8A CN201710776733A CN107545890A CN 107545890 A CN107545890 A CN 107545890A CN 201710776733 A CN201710776733 A CN 201710776733A CN 107545890 A CN107545890 A CN 107545890A
- Authority
- CN
- China
- Prior art keywords
- sound
- convolutional neural
- neural networks
- sound event
- cochlea
- 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.)
- Pending
Links
Landscapes
- Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)
Abstract
The present invention relates to a kind of sound event recognition method, mainly solve the problems, such as that the voice recognition accuracy rate of the prior art under powerful disturbed condition is low and poor robustness.By using following steps：Sound is acquired and handled under disturbance environment, forms audio digital signal；Sub-band filter is carried out to the audio digital signal by wave filter group, obtains the cochlea spectrogram of audio signal；A part for the cochlea spectrogram is trained to convolutional neural networks model, establishes voice recognition template；Another part of the cochlea spectrogram is substituted into the convolutional neural networks model, carries out the accuracy rate detection of the identification of sound；The above method preferably resolves the problem, the sound event identification that can be used under traffic environment.
Description
Technical field
The invention belongs to Audio Signal Processing technical field, is related specifically to a kind of sound event under strong interference environment
Recognition methods.
Background technology
In recent years, researchers propose many feature extracting methods and identifying system for sound event identification, all take
Certain effect was obtained, it is the acoustic information by gathering people to have a kind of sound event recognition method, and acoustic information is carried out at FFT
Reason, extracts the amplitude and frequency in everyone acoustic information, and store；The information of people is carried out after equally handling, with information
Information in storehouse is compared, and determines people's identity and carries out the sound event recognition method of identification, and this sound event is known
Other method recognition effect under small noise environment is preferable, but recognition effect is generally poor under very noisy, strong interference environment.
The content of the invention
The technical problems to be solved by the invention are that sound is known under very noisy, strong interference environment present in prior art
The poor technical problem of other effect, there is provided a kind of new sound event recognition method, the sound event recognition method have strong
Recognition accuracy and the high technical characterstic of robustness under noise, strong interference environment.
In order to solve the above technical problems, the technical scheme used is as follows：
A kind of sound event recognition method, comprises the following steps：
A. sound is acquired under interference environment, forms audio digital signal, the collection includes using sound level meter
Sound collection is carried out with microphone array；The processing is that the audio digital signal is carried out at end-point detection and filtering and noise reduction
Reason；
B. sub-band filter is carried out to the audio digital signal by wave filter group, obtains audio signal cochlea spectrogram；
C. a part for the cochlea spectrogram is trained to convolutional neural networks model, establishes sound event recognition template；
D. another part of the cochlea spectrogram is substituted into the convolutional neural networks model, carries out the identification of sound event
Accuracy rate detection.
In such scheme, for optimization, further, the extraction of the cochlea spectrogram comprises the following steps：
1) when audio digital signal described in is by the wave filter group, the expression formula of the response of the audio signal is exported
It is as follows：
Gm(i)=[| g | (i, m)]1/2, i=0,1 ..., N；M=0,1 ..., M-1
Wherein, Gm(i) matrix for representing changes in distribution on input audio signal frequency domain is formed, N is the audio signal
Port number, M are the frame number after sampling, obtain original cochlea spectrogram；
2) the original cochlea spectrogram is compressed, cutting obtains final cochlea spectrogram size, as the convolution
The input sample of neutral net.
Further, the method for building up of the sound event recognition template comprises the following steps：
I. using the cochlea spectrogram as learning sample, and the learning sample is done into class label；In the study sample
Learning sample of the part including all categories is extracted in this as training set, remaining part is as test set；
II. the convolutional neural networks model is built using software, the convolutional neural networks model includes setting gradually
The first convolutional layer, the first maximum pond layer, the second convolutional layer, the second maximum pond layer, full articulamentum and classification output layer；
III. the convolutional neural networks model is inputted using as the learning sample of the training set, exercise supervision study,
The parameter of each layer of the convolutional neural networks model after being trained；During training, using probability distribution function to convolution kernel
Random initializtion is carried out with weight, full 0 initialization is carried out to biasing；The algorithm adjustment weights and threshold declined using normal gradients
Value；Come training convolutional neural networks by way of network propagated forward and backpropagation repeatedly cross processing, until cost letter
Untill several limit errors is less than 0.01, the convolutional neural networks model trained is preserved；
IV. the convolutional neural networks model is tested, method of testing is：The test set sample is substituted into and trained
The good convolutional neural networks model, by the output of convolutional neural networks model sound corresponding with the test set sample
Sound classification is contrasted, and calculates the recall rate that sound event identifies under different signal to noise ratio, accuracy rate and F values respectively and to the volume
Product neural network model is assessed.
Further, the full articulamentum in step II is three, and the grader of the classification output layer is classified for softmax
Device.
Further, the sample of the training set in step I is the 3/4 of the learning sample.
Further, the wave filter group is formed for multiple Gammatone wave filters.
Further, the sound event of the step A collections and processing is included under traffic environment under different noise conditions
Vehicle collision sound, vehicle whistle sound, one or more sound events of personnel's sound of call for help or closing of the door sound.
Further, the audio digital signal carries out end-point detection using short-time energy double threshold thresholding algorithm.
Further, the audio digital signal is filtered denoising using LMP algorithms.
Further, first convolutional layer sets 20 convolution filters, and each wave filter size is 5 × 5, and convolution is moved
Dynamic step-length is 1, and activation primitive uses relu functions；The pond domain of the first maximum pond layer and the second maximum pond layer is
2 × 2, step-length is 2；Second convolutional layer sets 50 convolution filters, and each wave filter size is 5 × 5, convolution movement
Step-length is 1.
Compared with prior art, the beneficial effects of the invention are as follows：
1. the effect of the end-point detection of pair audio signal is that useful sound event information is extracted in very noisy or interference
Fragment；The effect of filtering and denoising is to reduce the influence of very noisy or interference to sound event feature extraction, to extract standard
True voice signal；Analog cochlea is simulated with wave filter group, signal frequency domain changes in distribution is described with obtained cochlea spectrogram,
Sound event during ambient noise can not only be detected or disturbed, and effective identification can be carried out to sound event and is divided
Class.
2. using the method for machine learning, manual intervention is avoided, convolutional neural networks model is fully learnt every class sound
The feature of sound event cochlea spectrogram, using convolutional neural networks generalization ability and adaptable characteristic, reaches higher identification
Accuracy rate and stronger robustness.
3. based on the sound event recognition method of convolutional neural networks model, there is preferable anti-noise ability, same noise
Under environment, discrimination of the invention is significantly improved.
4. the sound event recognition method of the present invention to be used for the traffic environment that complicated noise is high, interference is strong be present
Under, for vehicle collision sound, vehicle whistle sound, the sound event such as personnel's sound of call for help and closing of the door sound can have higher identification
Rate.
5. being filtered denoising to audio digital signal using LMP algorithms, effect is to reduce urban traffic noise to sound
The influence of affair character extraction.
6. using relu functions are used as activation primitive, the speed of training convolutional neural networks model can be improved.
7.Gammatone wave filters, which form wave filter group, can retain original sample frequency, be set on time dimension
After response frequency, extracted available for sound affair character in short-term.
Brief description of the drawings
The present invention is further described with reference to the accompanying drawings and examples.
Fig. 1 is the sound event identification process figure based on convolutional neural networks model.
Fig. 2 is the convolutional neural networks prototype network structure chart of voice recognition.
Description of reference numerals：
The maximum pond layer of maximum pond layer -3, the second convolutional layer -4, the second of cochlea spectrogram -1, the first convolutional layer -2, the first -
5, full articulamentum -6, output layer -7 of classifying.
Embodiment
In order to make the purpose , technical scheme and advantage of the present invention be clearer, with reference to embodiments, to the present invention
It is further elaborated.It should be appreciated that the specific embodiments described herein are merely illustrative of the present invention, it is not used to
Limit the present invention.
The present embodiment provides a kind of sound event recognition method, in the present embodiment using the noise conduct under traffic environment
Specific embodiment, idiographic flow is as shown in figure 1, comprise the following steps：
A. coordinate microphone array to be acquired sound using sound level meter under disturbance environment, form sound figure
Signal；
Respectively every kind of sound event is acquired and handles in 20dB, tri- kinds of signal to noise ratio of 10dB, 0dB to four kinds of sound events
Number of samples is 4800, sample frequency 8KHZ, and four kinds of sound events are vehicle collision sound, vehicle whistle sound, personnel's calling for help
The one or more of sound or closing of the door sound；
The voice data collected is pre-processed with matlab softwares.Utilize short-time energy double threshold thresholding algorithm pair
Audio digital signal carries out end-point detection, it is therefore an objective to useful sound event information segment is extracted in ambient noise, utilizes LMP
Algorithm is filtered denoising to audio digital signal, in order to reduces urban traffic noise to sound event feature extraction
Influence；
B. sub-band filter is carried out to the audio digital signal by wave filter group, obtains audio signal cochlea spectrogram；
The extracting method of cochlea spectrogram is as follows：
Analog cochlea is simulated using the 4 rank Gammatone wave filter groups of one group of 64 passage, realizes sub-band filter, its
Centre frequency is between 350Hz~4000Hz.Gammatone wave filters can retain original sample frequency, therefore in time dimension
It is 100Hz that response frequency is set on degree, the frame for producing 10ms is moved, available for sound feature extraction in short-term.When sound figure is believed
During number by Gammatone wave filter groups, the expression formula of the response of output signal is as follows：
Gm(i)=[| g | (i, m)]1/2, i=0,1 ..., N；M=0,1 ..., M-1
Wherein, Gm(i) matrix for representing changes in distribution on input audio signal frequency domain is formed, N is the passage of audio signal
Number, M are the frame number after sampling, and signal frequency domain changes in distribution is described using cochlea spectrogram；
Obtained original cochlea spectrogram is compressed, it is 32 × 32 that cutting, which obtains final cochlea spectrogram size, as
The input sample of convolutional neural networks；
C. a part for the cochlea spectrogram 1 is trained to convolutional neural networks model, i.e. CNN network structure models, built
Vertical sound event recognition template；
The method for building up of the convolutional neural networks model is as follows：
1) using the cochlea spectrogram of acquisition as learning sample, and class label is added to the learning sample；Different classes of
Learning sample in extract and 3/4 be used as training set, remaining 1/4 is test set；
2) the NVIDIA GTX1080 based on Pascal GP104 cores build training platform：Use MATLAB's
Parallel Computing Toolbox tool boxes and the establishment of Neural Network Toolbox tool boxes and training convolutional
Neural network model, model structure are as shown in Figure 2；
Determine the convolutional neural networks number of plies：Two convolutional layers, two pond layers and full articulamentum 6 and softmax graders
7, the full articulamentum 6 includes three full articulamentum 6-1,6-2,6-3, wherein, the first convolutional layer 2 sets 20 convolution filters,
Each wave filter size is 5 × 5, and convolution moving step length is 1, and to accelerate training speed, activation primitive uses relu functions；relu
Function is the linear unit function of amendment；The first maximum pond domain of pond layer 3 is 2 × 2, step-length 2；Second convolutional layer 4 sets 50
Individual convolution filter, each wave filter size are 5 × 5, and convolution moving step length is 1；The second maximum pond domain of pond layer 5 be 2 ×
2, step-length 2；Softmax graders 7 export four kinds of class objects：Vehicle collision sound, vehicle whistle sound, personnel's sound of call for help or car
The one or more of door closing sound.
3) training sample is inputted into convolutional neural networks, carries out the study for having supervision of tape label, the volume after being trained
The parameter of each layer of product neutral net.
In training process, random initializtion is carried out to convolution kernel and weight using probability distribution function, and biasing is carried out
Full 0 initializes.In order to accelerate algorithm adjustment weights and the threshold value that training process is then declined using normal gradients.By before network to
Propagate and the mode of backpropagation cross processing repeatedly carrys out training convolutional neural networks, until the limit error of cost function is less than
Untill 0.01, the convolutional neural networks model trained is preserved；
D. another part of the cochlea spectrogram is substituted into the convolutional neural networks model, carries out the identification of sound event
Accuracy rate detection；
The cochlea spectrogram of test set is substituted into the convolutional neural networks model trained, by the output and test of disaggregated model
Sound class corresponding to each cochlea spectrogram is concentrated to be contrasted, calculate that sound event under different signal to noise ratio identifies respectively recalls
Rate, accuracy rate and F values are assessed model.
Sound event number in the correct sound event number/sample for recall rate=extract；
Accuracy rate=the correct sound event number extracted/sound event number extracted；
F values=accuracy * recall rates * 2/ (accuracy+recall rate), F values are the harmonic average of accuracy and recall rate
Value.
Although the illustrative embodiment of the present invention is described above, in order to the technology of the art
Personnel are it will be appreciated that the present invention, but the present invention is not limited only to the scope of embodiment, to the common skill of the art
For art personnel, as long as long as various change in the spirit and scope of the invention that appended claim limits and determines, one
The innovation and creation using present inventive concept are cut in the row of protection.
Claims (10)
1. a kind of sound event recognition method, it is characterised in that comprise the following steps：
A. sound is acquired under interference environment, forms audio digital signal, the collection includes using sound level meter and wheat
Gram wind array carries out sound collection；The processing is that end-point detection and filtering and noise reduction processing are carried out to the audio digital signal；
B. sub-band filter is carried out to the audio digital signal by wave filter group, obtains audio signal cochlea spectrogram；
C. a part for the cochlea spectrogram is trained to convolutional neural networks model, establishes sound event recognition template；
D. another part of the cochlea spectrogram is substituted into the convolutional neural networks model, carries out the standard of the identification of sound event
True rate detection.
2. sound event recognition method according to claim 1, it is characterised in that：The extraction of the cochlea spectrogram include with
Lower step：
1) when the audio digital signal is by the wave filter group, the expression formula for exporting the response of the audio signal is as follows：
Gm(i)=[| g | (i, m)]1/2, i=0,1 ..., N；M=0,1 ..., M-1
Wherein, Gm(i) matrix for representing changes in distribution on input audio signal frequency domain is formed, N is the passage of the audio signal
Number, M are the frame number after sampling, obtain original cochlea spectrogram；
2) the original cochlea spectrogram is compressed, cutting obtains final cochlea spectrogram size, as the convolutional Neural
The input sample of network.
3. sound event recognition method according to claim 1, it is characterised in that：The foundation side of the voice recognition template
Method comprises the following steps：
I. using the cochlea spectrogram as learning sample, and class label is done to the learning sample；In the learning sample
Learning sample of the part including all categories is extracted as training set, remaining part is as test set；
II. build the convolutional neural networks model using software, the convolutional neural networks model include setting gradually the
One convolutional layer, the first maximum pond layer, the second convolutional layer, the second maximum pond layer, full articulamentum and classification output layer；
III. the convolutional neural networks model is inputted using as the learning sample of the training set, exercise supervision study, obtains
The parameter of each layer of the convolutional neural networks model after training；During training, using probability distribution function to convolution kernel and power
Random initializtion is carried out again, and full 0 initialization is carried out to biasing；The algorithm adjustment weights and threshold value declined using normal gradients；It is logical
The mode for crossing network propagated forward and backpropagation cross processing repeatedly carrys out training convolutional neural networks, until the limit of cost function
Untill determining error less than 0.01, the convolutional neural networks model trained is preserved；
IV. the convolutional neural networks model is tested, method of testing is：The sample of the test set is substituted into and trained
The convolutional neural networks model, by corresponding with the sample of the test set sound of output of the convolutional neural networks model
Sound classification is contrasted, and calculates the recall rate that sound event identifies under different signal to noise ratio, accuracy rate and F values respectively and to the volume
Product neural network model is assessed.
4. sound event recognition method according to claim 3, it is characterised in that：Full articulamentum in step II is three
Individual, the grader of the classification output layer is softmax graders.
5. sound event recognition method according to claim 4, it is characterised in that：The sample of training set in step I is
The 3/4 of the learning sample.
6. according to any described sound event recognition methods of claim 1-5, it is characterised in that：The wave filter group is more logical
Road Gammatone wave filters are formed.
7. sound event recognition method according to claim 6, it is characterised in that：The sound of the step A collections and processing
Vehicle collision sound, vehicle whistle sound, personnel's sound of call for help or the car door that sound event includes under traffic environment under different noise conditions close
Close one or more sound events of sound.
8. sound event recognition method according to claim 7, it is characterised in that：The audio digital signal is using in short-term
Energy double threshold thresholding algorithm carries out end-point detection.
9. sound event recognition method according to claim 7, it is characterised in that：The audio digital signal uses LMP
Algorithm is filtered and denoising.
10. sound event recognition method according to claim 8, it is characterised in that：First convolutional layer sets 20
Convolution filter, each wave filter size are 5 × 5, and convolution moving step length is 1, and activation primitive uses relu functions；Described first
The pond domain of maximum pond layer and the second maximum pond layer is 2 × 2, and step-length is 2；Second convolutional layer sets 50 volumes
Product wave filter, each wave filter size are 5 × 5, and convolution moving step length is 1.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710776733.8A CN107545890A (en) | 2017-08-31 | 2017-08-31 | A kind of sound event recognition method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710776733.8A CN107545890A (en) | 2017-08-31 | 2017-08-31 | A kind of sound event recognition method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN107545890A true CN107545890A (en) | 2018-01-05 |
Family
ID=60959363
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710776733.8A Pending CN107545890A (en) | 2017-08-31 | 2017-08-31 | A kind of sound event recognition method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107545890A (en) |
Cited By (43)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108172238A (en) * | 2018-01-06 | 2018-06-15 | 广州音书科技有限公司 | A kind of voice enhancement algorithm based on multiple convolutional neural networks in speech recognition system |
| CN108519149A (en) * | 2018-03-28 | 2018-09-11 | 长安大学 | A kind of tunnel accident monitor and alarm system and method based on sound Time-Frequency Analysis |
| CN108538311A (en) * | 2018-04-13 | 2018-09-14 | 腾讯音乐娱乐科技（深圳）有限公司 | Audio frequency classification method, device and computer readable storage medium |
| CN108615533A (en) * | 2018-03-28 | 2018-10-02 | 天津大学 | A kind of high-performance sound enhancement method based on deep learning |
| CN108806698A (en) * | 2018-03-15 | 2018-11-13 | 中山大学 | A kind of camouflage audio recognition method based on convolutional neural networks |
| CN108854062A (en) * | 2018-06-24 | 2018-11-23 | 广州银汉科技有限公司 | A kind of voice-enabled chat module of moving game |
| CN108932480A (en) * | 2018-06-08 | 2018-12-04 | 电子科技大学 | The study of distributing optical fiber sensing signal characteristic and classification method based on 1D-CNN |
| CN109001557A (en) * | 2018-06-11 | 2018-12-14 | 西北工业大学 | A kind of aircraft utilities system fault recognition method based on random convolutional neural networks |
| CN109065030A (en) * | 2018-08-01 | 2018-12-21 | 上海大学 | Ambient sound recognition methods and system based on convolutional neural networks |
| CN109102798A (en) * | 2018-06-29 | 2018-12-28 | 厦门快商通信息技术有限公司 | A kind of finishing event detecting method, device, computer equipment and medium |
| CN109243496A (en) * | 2018-10-31 | 2019-01-18 | 东方智测（北京）科技有限公司 | Sound identification method and system |
| CN109348086A (en) * | 2018-11-05 | 2019-02-15 | 重庆大学 | The high-effect image synchronization identification of intelligent radio video camera and compression method |
| CN109376264A (en) * | 2018-11-09 | 2019-02-22 | 广州势必可赢网络科技有限公司 | A kind of audio-frequency detection, device, equipment and computer readable storage medium |
| CN109472311A (en) * | 2018-11-13 | 2019-03-15 | 北京物灵智能科技有限公司 | A kind of user behavior recognition method and device |
| CN109473120A (en) * | 2018-11-14 | 2019-03-15 | 辽宁工程技术大学 | A kind of abnormal sound signal recognition method based on convolutional neural networks |
| CN109631104A (en) * | 2018-11-01 | 2019-04-16 | 广东万和热能科技有限公司 | Air quantity Automatic adjustment method, device, equipment and the storage medium of kitchen ventilator |
| CN109800860A (en) * | 2018-12-28 | 2019-05-24 | 北京工业大学 | A kind of Falls in Old People detection method of the Community-oriented based on CNN algorithm |
| CN109986409A (en) * | 2019-04-11 | 2019-07-09 | 中国一拖集团有限公司 | A kind of flutter recognition methods and system based on convolutional neural networks |
| CN110047512A (en) * | 2019-04-25 | 2019-07-23 | 广东工业大学 | A kind of ambient sound classification method, system and relevant apparatus |
| CN110046655A (en) * | 2019-03-26 | 2019-07-23 | 天津大学 | A kind of audio scene recognition method based on integrated study |
| CN110136745A (en) * | 2019-05-08 | 2019-08-16 | 西北工业大学 | A kind of vehicle whistle recognition methods based on convolutional neural networks |
| CN110176248A (en) * | 2019-05-23 | 2019-08-27 | 广西交通科学研究院有限公司 | Road sound identification method, system, computer equipment and readable storage medium storing program for executing |
| WO2020000523A1 (en) * | 2018-06-26 | 2020-01-02 | 深圳大学 | Signal processing method and apparatus |
| CN110808067A (en) * | 2019-11-08 | 2020-02-18 | 福州大学 | Low signal-to-noise ratio sound event detection method based on binary multiband energy distribution |
| CN110824006A (en) * | 2019-11-08 | 2020-02-21 | 南通大学 | Postweld weld impact quality discrimination method based on intelligent acoustic information identification |
| CN110942670A (en) * | 2019-11-20 | 2020-03-31 | 神思电子技术股份有限公司 | Expressway fog area induction method |
| CN110992979A (en) * | 2019-11-29 | 2020-04-10 | 北京搜狗科技发展有限公司 | Detection method and device and electronic equipment |
| CN111445926A (en) * | 2020-04-01 | 2020-07-24 | 杭州叙简科技股份有限公司 | Rural road traffic accident warning condition identification method based on sound |
| CN111476102A (en) * | 2020-03-11 | 2020-07-31 | 华中科技大学鄂州工业技术研究院 | Safety protection method, central control equipment and computer storage medium |
| CN111524536A (en) * | 2019-02-01 | 2020-08-11 | 富士通株式会社 | Signal processing method and information processing apparatus |
| CN111627419A (en) * | 2020-05-09 | 2020-09-04 | 哈尔滨工程大学 | Sound generation method based on underwater target and environmental information characteristics |
| CN111653067A (en) * | 2020-06-12 | 2020-09-11 | 杭州海康威视数字技术股份有限公司 | Intelligent household equipment and alarm method based on audio frequency |
| CN111699368A (en) * | 2019-05-22 | 2020-09-22 | 深圳市大疆创新科技有限公司 | Strike detection method, device, movable platform and computer readable storage medium |
| CN112086105A (en) * | 2020-08-31 | 2020-12-15 | 中国船舶重工集团公司七五0试验场 | Target identification method based on Gamma atom sub-band continuous spectrum characteristics |
| CN112349298A (en) * | 2019-08-09 | 2021-02-09 | 阿里巴巴集团控股有限公司 | Sound event recognition method, device, equipment and storage medium |
| CN112363114A (en) * | 2021-01-14 | 2021-02-12 | 杭州兆华电子有限公司 | Public place acoustic event positioning method and system based on distributed noise sensor |
| CN112419258A (en) * | 2020-11-18 | 2021-02-26 | 西北工业大学 | Robust environmental sound identification method based on time-frequency segmentation and convolutional neural network |
| CN112447187A (en) * | 2019-09-02 | 2021-03-05 | 富士通株式会社 | Device and method for recognizing sound event |
| CN112633227A (en) * | 2020-12-30 | 2021-04-09 | 应急管理部国家自然灾害防治研究院 | Automatic identification method and system for Zhang Heng I induction magnetometer data lightning whistle sound wave |
| CN113112681A (en) * | 2020-01-13 | 2021-07-13 | 阿里健康信息技术有限公司 | Vending equipment, and shipment detection method and device |
| CN114722884A (en) * | 2022-06-08 | 2022-07-08 | 深圳市润东来科技有限公司 | Audio control method, device and equipment based on environmental sound and storage medium |
| CN116052452A (en) * | 2023-04-03 | 2023-05-02 | 江西方兴科技股份有限公司 | Data processing method and lane early warning method for wireless communication |
| CN116506786A (en) * | 2023-05-12 | 2023-07-28 | 深圳市英唐数码科技有限公司 | Hearing-aid device performance intelligent monitoring method, system and medium |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20160284347A1 (en) * | 2015-03-27 | 2016-09-29 | Google Inc. | Processing audio waveforms |
| CN106328150A (en) * | 2016-08-18 | 2017-01-11 | 北京易迈医疗科技有限公司 | Bowel sound detection method, device and system under noisy environment |
| CN106682574A (en) * | 2016-11-18 | 2017-05-17 | 哈尔滨工程大学 | One-dimensional deep convolution network underwater multi-target recognition method |
| CN106846803A (en) * | 2017-02-08 | 2017-06-13 | 广西交通科学研究院有限公司 | Traffic incidents detection device and method based on audio |
-
2017
- 2017-08-31 CN CN201710776733.8A patent/CN107545890A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20160284347A1 (en) * | 2015-03-27 | 2016-09-29 | Google Inc. | Processing audio waveforms |
| CN106328150A (en) * | 2016-08-18 | 2017-01-11 | 北京易迈医疗科技有限公司 | Bowel sound detection method, device and system under noisy environment |
| CN106682574A (en) * | 2016-11-18 | 2017-05-17 | 哈尔滨工程大学 | One-dimensional deep convolution network underwater multi-target recognition method |
| CN106846803A (en) * | 2017-02-08 | 2017-06-13 | 广西交通科学研究院有限公司 | Traffic incidents detection device and method based on audio |
Cited By (59)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108172238A (en) * | 2018-01-06 | 2018-06-15 | 广州音书科技有限公司 | A kind of voice enhancement algorithm based on multiple convolutional neural networks in speech recognition system |
| CN108172238B (en) * | 2018-01-06 | 2021-08-13 | 广州音书科技有限公司 | Speech enhancement algorithm based on multiple convolutional neural networks in speech recognition system |
| CN108806698A (en) * | 2018-03-15 | 2018-11-13 | 中山大学 | A kind of camouflage audio recognition method based on convolutional neural networks |
| CN108519149A (en) * | 2018-03-28 | 2018-09-11 | 长安大学 | A kind of tunnel accident monitor and alarm system and method based on sound Time-Frequency Analysis |
| CN108615533A (en) * | 2018-03-28 | 2018-10-02 | 天津大学 | A kind of high-performance sound enhancement method based on deep learning |
| CN108538311A (en) * | 2018-04-13 | 2018-09-14 | 腾讯音乐娱乐科技（深圳）有限公司 | Audio frequency classification method, device and computer readable storage medium |
| CN108932480A (en) * | 2018-06-08 | 2018-12-04 | 电子科技大学 | The study of distributing optical fiber sensing signal characteristic and classification method based on 1D-CNN |
| CN108932480B (en) * | 2018-06-08 | 2022-03-15 | 电子科技大学 | Distributed optical fiber sensing signal feature learning and classifying method based on 1D-CNN |
| CN109001557A (en) * | 2018-06-11 | 2018-12-14 | 西北工业大学 | A kind of aircraft utilities system fault recognition method based on random convolutional neural networks |
| CN108854062A (en) * | 2018-06-24 | 2018-11-23 | 广州银汉科技有限公司 | A kind of voice-enabled chat module of moving game |
| WO2020000523A1 (en) * | 2018-06-26 | 2020-01-02 | 深圳大学 | Signal processing method and apparatus |
| CN109102798A (en) * | 2018-06-29 | 2018-12-28 | 厦门快商通信息技术有限公司 | A kind of finishing event detecting method, device, computer equipment and medium |
| CN109065030A (en) * | 2018-08-01 | 2018-12-21 | 上海大学 | Ambient sound recognition methods and system based on convolutional neural networks |
| CN109065030B (en) * | 2018-08-01 | 2020-06-30 | 上海大学 | Convolutional neural network-based environmental sound identification method and system |
| CN109243496A (en) * | 2018-10-31 | 2019-01-18 | 东方智测（北京）科技有限公司 | Sound identification method and system |
| CN109631104A (en) * | 2018-11-01 | 2019-04-16 | 广东万和热能科技有限公司 | Air quantity Automatic adjustment method, device, equipment and the storage medium of kitchen ventilator |
| CN109348086B (en) * | 2018-11-05 | 2020-09-15 | 重庆大学 | Intelligent wireless camera image synchronous identification and compression method |
| CN109348086A (en) * | 2018-11-05 | 2019-02-15 | 重庆大学 | The high-effect image synchronization identification of intelligent radio video camera and compression method |
| CN109376264A (en) * | 2018-11-09 | 2019-02-22 | 广州势必可赢网络科技有限公司 | A kind of audio-frequency detection, device, equipment and computer readable storage medium |
| CN109472311A (en) * | 2018-11-13 | 2019-03-15 | 北京物灵智能科技有限公司 | A kind of user behavior recognition method and device |
| CN109473120A (en) * | 2018-11-14 | 2019-03-15 | 辽宁工程技术大学 | A kind of abnormal sound signal recognition method based on convolutional neural networks |
| CN109800860A (en) * | 2018-12-28 | 2019-05-24 | 北京工业大学 | A kind of Falls in Old People detection method of the Community-oriented based on CNN algorithm |
| CN111524536B (en) * | 2019-02-01 | 2023-09-08 | 富士通株式会社 | Signal processing method and information processing apparatus |
| CN111524536A (en) * | 2019-02-01 | 2020-08-11 | 富士通株式会社 | Signal processing method and information processing apparatus |
| CN110046655B (en) * | 2019-03-26 | 2023-03-31 | 天津大学 | Audio scene recognition method based on ensemble learning |
| CN110046655A (en) * | 2019-03-26 | 2019-07-23 | 天津大学 | A kind of audio scene recognition method based on integrated study |
| CN109986409A (en) * | 2019-04-11 | 2019-07-09 | 中国一拖集团有限公司 | A kind of flutter recognition methods and system based on convolutional neural networks |
| CN110047512A (en) * | 2019-04-25 | 2019-07-23 | 广东工业大学 | A kind of ambient sound classification method, system and relevant apparatus |
| CN110136745A (en) * | 2019-05-08 | 2019-08-16 | 西北工业大学 | A kind of vehicle whistle recognition methods based on convolutional neural networks |
| CN111699368A (en) * | 2019-05-22 | 2020-09-22 | 深圳市大疆创新科技有限公司 | Strike detection method, device, movable platform and computer readable storage medium |
| CN110176248A (en) * | 2019-05-23 | 2019-08-27 | 广西交通科学研究院有限公司 | Road sound identification method, system, computer equipment and readable storage medium storing program for executing |
| CN112349298A (en) * | 2019-08-09 | 2021-02-09 | 阿里巴巴集团控股有限公司 | Sound event recognition method, device, equipment and storage medium |
| CN112447187A (en) * | 2019-09-02 | 2021-03-05 | 富士通株式会社 | Device and method for recognizing sound event |
| WO2021088176A1 (en) * | 2019-11-08 | 2021-05-14 | 福州大学 | Binary multi-band power distribution-based low signal-to-noise ratio sound event detection method |
| CN110824006B (en) * | 2019-11-08 | 2021-12-28 | 南通大学 | Postweld weld impact quality discrimination method based on intelligent acoustic information identification |
| CN110808067A (en) * | 2019-11-08 | 2020-02-18 | 福州大学 | Low signal-to-noise ratio sound event detection method based on binary multiband energy distribution |
| CN110824006A (en) * | 2019-11-08 | 2020-02-21 | 南通大学 | Postweld weld impact quality discrimination method based on intelligent acoustic information identification |
| CN110942670A (en) * | 2019-11-20 | 2020-03-31 | 神思电子技术股份有限公司 | Expressway fog area induction method |
| CN110992979B (en) * | 2019-11-29 | 2022-04-08 | 北京搜狗科技发展有限公司 | Detection method and device and electronic equipment |
| CN110992979A (en) * | 2019-11-29 | 2020-04-10 | 北京搜狗科技发展有限公司 | Detection method and device and electronic equipment |
| CN113112681A (en) * | 2020-01-13 | 2021-07-13 | 阿里健康信息技术有限公司 | Vending equipment, and shipment detection method and device |
| CN111476102A (en) * | 2020-03-11 | 2020-07-31 | 华中科技大学鄂州工业技术研究院 | Safety protection method, central control equipment and computer storage medium |
| CN111445926B (en) * | 2020-04-01 | 2023-01-03 | 杭州叙简科技股份有限公司 | Rural road traffic accident warning condition identification method based on sound |
| CN111445926A (en) * | 2020-04-01 | 2020-07-24 | 杭州叙简科技股份有限公司 | Rural road traffic accident warning condition identification method based on sound |
| CN111627419B (en) * | 2020-05-09 | 2022-03-22 | 哈尔滨工程大学 | Sound generation method based on underwater target and environmental information characteristics |
| CN111627419A (en) * | 2020-05-09 | 2020-09-04 | 哈尔滨工程大学 | Sound generation method based on underwater target and environmental information characteristics |
| CN111653067A (en) * | 2020-06-12 | 2020-09-11 | 杭州海康威视数字技术股份有限公司 | Intelligent household equipment and alarm method based on audio frequency |
| CN112086105A (en) * | 2020-08-31 | 2020-12-15 | 中国船舶重工集团公司七五0试验场 | Target identification method based on Gamma atom sub-band continuous spectrum characteristics |
| CN112086105B (en) * | 2020-08-31 | 2022-08-19 | 中国船舶重工集团公司七五0试验场 | Target identification method based on Gamma atom sub-band continuous spectrum characteristics |
| CN112419258A (en) * | 2020-11-18 | 2021-02-26 | 西北工业大学 | Robust environmental sound identification method based on time-frequency segmentation and convolutional neural network |
| CN112419258B (en) * | 2020-11-18 | 2024-05-14 | 西北工业大学 | Robust environment sound identification method based on time-frequency segmentation and convolutional neural network |
| CN112633227A (en) * | 2020-12-30 | 2021-04-09 | 应急管理部国家自然灾害防治研究院 | Automatic identification method and system for Zhang Heng I induction magnetometer data lightning whistle sound wave |
| CN112633227B (en) * | 2020-12-30 | 2024-02-23 | 应急管理部国家自然灾害防治研究院 | Automatic recognition method and system for data lightning whistle sound waves of Zhangheng first induction magnetometer |
| CN112363114A (en) * | 2021-01-14 | 2021-02-12 | 杭州兆华电子有限公司 | Public place acoustic event positioning method and system based on distributed noise sensor |
| CN114722884A (en) * | 2022-06-08 | 2022-07-08 | 深圳市润东来科技有限公司 | Audio control method, device and equipment based on environmental sound and storage medium |
| CN114722884B (en) * | 2022-06-08 | 2022-09-30 | 深圳市润东来科技有限公司 | Audio control method, device and equipment based on environmental sound and storage medium |
| CN116052452A (en) * | 2023-04-03 | 2023-05-02 | 江西方兴科技股份有限公司 | Data processing method and lane early warning method for wireless communication |
| CN116506786A (en) * | 2023-05-12 | 2023-07-28 | 深圳市英唐数码科技有限公司 | Hearing-aid device performance intelligent monitoring method, system and medium |
| CN116506786B (en) * | 2023-05-12 | 2024-05-31 | 深圳市英唐数码科技有限公司 | Hearing-aid device performance intelligent monitoring method, system and medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107545890A (en) | A kind of sound event recognition method | |
| CN109767785A (en) | Ambient noise method for identifying and classifying based on convolutional neural networks | |
| CN113707176B (en) | Transformer fault detection method based on acoustic signal and deep learning technology | |
| CN110767216B (en) | Voice recognition attack defense method based on PSO algorithm | |
| CN108680245A (en) | Whale globefish class Click classes are called and traditional Sonar Signal sorting technique and device | |
| CN109473120A (en) | A kind of abnormal sound signal recognition method based on convolutional neural networks | |
| CN106847309A (en) | A kind of speech-emotion recognition method | |
| CN106710599A (en) | Particular sound source detection method and particular sound source detection system based on deep neural network | |
| CN107393542A (en) | A kind of birds species identification method based on binary channels neutral net | |
| CN106846803A (en) | Traffic incidents detection device and method based on audio | |
| CN110428843A (en) | A kind of voice gender identification deep learning method | |
| CN107993648A (en) | A kind of unmanned plane recognition methods, device and electronic equipment | |
| CN106847293A (en) | Facility cultivation sheep stress behavior acoustical signal monitoring method | |
| CN112735473B (en) | Method and system for identifying unmanned aerial vehicle based on voice | |
| CN113724712B (en) | Bird sound identification method based on multi-feature fusion and combination model | |
| CN105424366A (en) | Bearing fault diagnosis method based on EEMD adaptive denoising | |
| CN102779510A (en) | Speech emotion recognition method based on feature space self-adaptive projection | |
| CN113566948A (en) | Fault audio recognition and diagnosis method for robot coal pulverizer | |
| CN111580151A (en) | SSNet model-based earthquake event time-of-arrival identification method | |
| CN103994820B (en) | A kind of moving target recognition methods based on micropore diameter microphone array | |
| CN110321810A (en) | Single channel signal two-way separation method, device, storage medium and processor | |
| CN109631104A (en) | Air quantity Automatic adjustment method, device, equipment and the storage medium of kitchen ventilator | |
| CN106548786A (en) | A kind of detection method and system of voice data | |
| CN115081473A (en) | Multi-feature fusion brake noise classification and identification method | |
| CN112133323A (en) | Unsupervised classification and supervised modification fusion voice separation method related to spatial structural characteristics |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180105 |
|
| RJ01 | Rejection of invention patent application after publication |