CN109360584A - Cough monitoring method and device based on deep learning - Google Patents
Cough monitoring method and device based on deep learning Download PDFInfo
- Publication number
- CN109360584A CN109360584A CN201811259953.4A CN201811259953A CN109360584A CN 109360584 A CN109360584 A CN 109360584A CN 201811259953 A CN201811259953 A CN 201811259953A CN 109360584 A CN109360584 A CN 109360584A
- Authority
- CN
- China
- Prior art keywords
- cough
- audio
- audio data
- data
- sounder
- 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
- 206010011224 Cough Diseases 0.000 title claims abstract description 380
- 238000012544 monitoring process Methods 0.000 title claims abstract description 122
- 238000000034 method Methods 0.000 title claims abstract description 45
- 238000013135 deep learning Methods 0.000 title claims abstract description 28
- 230000001256 tonic effect Effects 0.000 claims abstract description 44
- 238000012545 processing Methods 0.000 claims abstract description 20
- 239000013598 vector Substances 0.000 claims description 27
- 230000015654 memory Effects 0.000 claims description 13
- 230000001755 vocal effect Effects 0.000 claims description 12
- 238000001514 detection method Methods 0.000 claims description 11
- 238000004590 computer program Methods 0.000 claims description 9
- 238000012806 monitoring device Methods 0.000 claims description 8
- 238000007781 pre-processing Methods 0.000 claims description 8
- 230000011218 segmentation Effects 0.000 claims description 5
- 238000013473 artificial intelligence Methods 0.000 abstract description 2
- 210000002569 neuron Anatomy 0.000 description 14
- 238000010586 diagram Methods 0.000 description 10
- 238000000605 extraction Methods 0.000 description 6
- 230000008569 process Effects 0.000 description 6
- 238000012549 training Methods 0.000 description 6
- 238000013528 artificial neural network Methods 0.000 description 4
- 230000007246 mechanism Effects 0.000 description 4
- 230000001537 neural effect Effects 0.000 description 4
- 230000002457 bidirectional effect Effects 0.000 description 3
- 238000004891 communication Methods 0.000 description 3
- 230000006870 function Effects 0.000 description 3
- 238000003062 neural network model Methods 0.000 description 3
- 238000001228 spectrum Methods 0.000 description 3
- 230000004913 activation Effects 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 2
- 238000013075 data extraction Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000011084 recovery Methods 0.000 description 2
- 238000012546 transfer Methods 0.000 description 2
- 230000009466 transformation Effects 0.000 description 2
- 241001269238 Data Species 0.000 description 1
- 208000000059 Dyspnea Diseases 0.000 description 1
- 206010013975 Dyspnoeas Diseases 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 230000006835 compression Effects 0.000 description 1
- 238000007906 compression Methods 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 1
- 208000017574 dry cough Diseases 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000004807 localization Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000000306 recurrent effect Effects 0.000 description 1
- 238000004064 recycling Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 208000023504 respiratory system disease Diseases 0.000 description 1
- 230000006403 short-term memory Effects 0.000 description 1
- 230000002123 temporal effect Effects 0.000 description 1
- 238000013519 translation Methods 0.000 description 1
- 230000001960 triggered effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L25/00—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
- G10L25/48—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 specially adapted for particular use
- G10L25/51—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 specially adapted for particular use for comparison or discrimination
- G10L25/66—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 specially adapted for particular use for comparison or discrimination for extracting parameters related to health condition
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L17/00—Speaker identification or verification techniques
- G10L17/02—Preprocessing operations, e.g. segment selection; Pattern representation or modelling, e.g. based on linear discriminant analysis [LDA] or principal components; Feature selection or extraction
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L25/00—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
- G10L25/03—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the type of extracted parameters
- G10L25/18—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the type of extracted parameters the extracted parameters being spectral information of each sub-band
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Multimedia (AREA)
- Acoustics & Sound (AREA)
- Human Computer Interaction (AREA)
- Signal Processing (AREA)
- Computational Linguistics (AREA)
- Public Health (AREA)
- General Health & Medical Sciences (AREA)
- Epidemiology (AREA)
- Spectroscopy & Molecular Physics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)
Abstract
This disclosure relates to which field of artificial intelligence, discloses a kind of cough monitoring method and device based on deep learning, comprising: pre-processed to obtain several frame tonic trains to the audio data of acquisition；Cough identification is carried out to several frame tonic trains, whether is cough audio with the determination audio data；And Application on Voiceprint Recognition is carried out to several frame tonic trains, with the sounder of the determination audio data；If the audio data is cough audio, processing is updated according to cough monitoring data of the audio data to the sounder.Cough identification and Application on Voiceprint Recognition are carried out to audio data using the method for deep learning, to obtain the cough monitoring data that audio data corresponds to sounder, realize the automatic monitoring to cough, it is convenient and efficient, it does not need special messenger and carries out cough monitoring, the efficiency for improving cough monitoring ensure that the real-time of cough monitoring data.
Description
Technical field
This disclosure relates to field of artificial intelligence, in particular to a kind of cough monitoring method and dress based on deep learning
It sets.
Background technique
Cough is the common sympton of respiratory disease, by the status of cough of patient can reflect patient physical condition or
The recovery situation of patient's person state of an illness.The status of cough of general patient is entourage's (such as families of patients) by patient in day
Often accompany and attend to what Cough length recorded in process, cough number etc. were got.But entourage is not to accompany at any time
At one's side, thus by entourage, that status of cough obtained is recorded during accompanying and attending to is not comprehensive, and by special by patient
Entourage it is also big come the status of cough workload for recording patient.
Therefore cough monitoring effect and cough monitoring efficiency need to be improved.
Summary of the invention
In order to solve the problems, such as present in the relevant technologies, present disclose provides a kind of cough monitoring side based on deep learning
Method and device.
A kind of cough monitoring method based on deep learning, comprising:
The audio data of acquisition is pre-processed to obtain several frame tonic trains；
Cough identification is carried out to several frame tonic trains, whether is cough audio with the determination audio data；With
And
Application on Voiceprint Recognition is carried out to several frame tonic trains, with the sounder of the determination audio data；
If the audio data is cough audio, number is monitored to the cough of the sounder according to the audio data
It is handled according to being updated.
A kind of cough monitoring device based on deep learning, comprising:
Preprocessing module is configured as executing: being pre-processed to obtain several frame tonic trains to the audio data of acquisition；
Cough identification module, is configured as executing: cough identification is carried out to several frame tonic trains, described in determination
Whether audio data is cough audio；And
Voiceprint identification module is configured as executing: Application on Voiceprint Recognition is carried out to several frame tonic trains, described in determination
The sounder of audio data；
Monitoring data update module is configured as executing: if the audio data is cough audio, according to the sound
Frequency evidence is updated processing to the cough monitoring data of the sounder.
In one embodiment, the preprocessing module includes:
Short Time Fourier Transform unit is configured as executing: carrying out Short Time Fourier Transform to the audio data of acquisition and obtains
To the corresponding sonograph of the audio data；
Segmenting unit is configured as executing: carrying out the segmentation of the sonograph according to specified duration, obtains several frames
Tonic train.
In one embodiment, the cough identification module includes:
Local feature vectors extraction unit is configured as executing: is extracted described in obtaining from several frame tonic trains
Several local feature vectors of audio data；
Full connection unit is configured as executing: carrying out the full connection of several local feature vectors, obtains the audio
The global characteristics vector of data；
Cough label prediction unit, is configured as executing: carrying out cough Tag Estimation to the global characteristics vector, obtain
The cough label of the audio data；
Interpretation unit is configured as executing: judging whether the audio data is cough audio according to the cough label.
In one embodiment, the voiceprint identification module includes:
Vocal print feature extraction unit is configured as executing: extracting from several frame tonic trains and obtains the audio
The vocal print feature of data；
Sounder Tag Estimation module, is configured as executing: carrying out sounder Tag Estimation according to the vocal print feature, obtains
To the sounder label of the audio data；
Sounder determining module is configured as executing: the sounding of the audio data is determined according to the sounder label
Person.
In one embodiment, monitoring data update module includes:
Timestamp information extraction unit is configured as executing: if the audio data is cough audio, from the sound
Frequency is extracted in and obtains timestamp information；
Cough length information and cough frequency information generating unit, are configured as executing: according to the cough audio and
The corresponding timestamp information generates Cough length information and cough frequency information；
Cough monitoring data acquiring unit in short-term, is configured as executing: by the Cough length information and cough frequency letter
Cease the cough monitoring data in short-term as the sounder in the audio data；
Monitoring data updating unit is configured as executing: the monitoring data of cough in short-term are uploaded to monitor database
In, to update the cough monitoring data of sounder described in the monitor database.
In one embodiment, Cough length information and cough frequency information generating unit include:
End-point detection unit is configured as executing: carrying out end-point detection to the cough audio with the determination cough sound
The beginning and end coughed every time in frequency；
Acquiring unit is configured as executing: being obtained according to timestamp information corresponding to the beginning and end coughed every time
The points or terminal number coughed into the Cough length information, and the statistics cough audio obtain the cough frequency
Information.
A kind of cough monitoring device based on deep learning, comprising:
Processor；And
Memory for storage processor executable instruction；
Wherein, the above-described cough based on deep learning is realized when the executable instruction is executed by the processor
Monitoring method.
A kind of computer readable storage medium, is stored thereon with computer program, and the computer program is held by processor
The above-described cough monitoring method based on deep learning is realized when row.
The technical scheme provided by this disclosed embodiment can include the following benefits: by carrying out cough to audio data
It coughs identification and Application on Voiceprint Recognition, to determine whether audio data is cough audio and the corresponding sounder of audio data, in turn
The cough monitoring data of sounder are obtained to cough audio analysis, realize the automatic monitoring to cough, it is convenient and efficient, it does not need
Special messenger carries out cough monitoring, improves the efficiency of cough monitoring, ensure that the real-time of cough monitoring data.
It should be understood that the above general description and the following detailed description are merely exemplary, this can not be limited
It is open.
Detailed description of the invention
The drawings herein are incorporated into the specification and forms part of this specification, and shows and meets implementation of the invention
Example, and in specification together principle for explaining the present invention.
Fig. 1 is the schematic diagram of the implementation environment according to involved in the disclosure；
Fig. 2 is a kind of block diagram of monitoring server of coughing shown according to an exemplary embodiment；
Fig. 3 is a kind of flow chart of cough monitoring method based on deep learning shown according to an exemplary embodiment；
Fig. 4 is the flow chart of the step S110 of embodiment illustrated in fig. 3；
Fig. 5 is the flow chart of the step S130 of embodiment illustrated in fig. 3；
Fig. 6 is the flow chart of the step S140 of embodiment illustrated in fig. 3；
Fig. 7 is the flow chart of the step S150 of embodiment illustrated in fig. 3；
Fig. 8 is the flow chart of the step S152 of embodiment illustrated in fig. 7；
Fig. 9 is the flow chart after the step S150 of embodiment illustrated in fig. 3；
Figure 10 is the schematic diagram for the neural network model for identification of coughing；
Figure 11 is the block diagram of the cough monitoring device shown according to an exemplary embodiment based on deep learning；
Figure 12 is the block diagram of the cough monitoring device based on deep learning shown according to another exemplary embodiment.
Specific embodiment
Here will the description is performed on the exemplary embodiment in detail, the example is illustrated in the accompanying drawings.Following description is related to
When attached drawing, unless otherwise indicated, the same numbers in different drawings indicate the same or similar elements.Following exemplary embodiment
Described in embodiment do not represent all embodiments consistented with the present invention.On the contrary, they be only with it is such as appended
The example of device and method being described in detail in claims, some aspects of the invention are consistent.
Fig. 1 is the schematic diagram of the implementation environment according to involved in the disclosure.The implementation environment includes: cough monitoring server
200.Cough monitoring server 200 can carry out cough identification and Application on Voiceprint Recognition to audio data using the method that the disclosure provides
To obtain cough monitoring data.
As needed, in the implementation environment will also include audio collecting device 100, for provide cough monitoring server into
The audio data of row cough identification and Application on Voiceprint Recognition.Audio collecting device 100 can be arranged in the living environment of related sounder
In, such as family, for the application scenarios of hospital, audio collecting device 100 can be arranged in the ward of patient (for this kind
Situation, patient are sounder involved in the disclosure.) the implementation progress audio data collecting of audio collecting device 100, then will
Collected audio data is transferred to cough monitoring server 200 and is handled, and then obtains the cough audio data of sounder.
Wherein audio collecting device 100 can be independent audio collecting device, such as microphone, recorder, can be with
Electronic equipment of microphone, such as video recorder, smart phone, computer etc. are carried, is not defined herein.
As needed, in the implementation environment will also include cough monitoring terminal 300, the cough monitor terminal 300 be used for into
The display etc. of row cough monitoring data.The cough monitoring data for the sounder that cough monitoring server 200 obtains processing return
Into cough monitoring terminal 300, consequently facilitating user knows the cough monitoring data of implementation.
Wireless or wired network is pre-established between audio collecting device 100 and cough monitoring server 200 to connect
It connects, and then is transmitted by the data that network connection realizes audio collecting device 100 between monitoring server 200 of coughing, for example,
The audio data that audio collecting device 100 acquires.Correspondingly, pre- between cough monitoring terminal 300 and cough monitoring server 200
Wired or wireless network connection is first established, and then cough monitoring server 200 and cough can be realized by network connection
The data monitored between terminal 300 of coughing are transmitted.
It should be noted that cough monitoring method of the disclosure based on deep learning, be not limited in cough monitoring server
Corresponding processing logic is disposed in 200, is also possible to the processing logic being deployed in other machines.For example, being calculated having
Deployment carries out the processing logic of cough identification and Application on Voiceprint Recognition to the audio data of acquisition in the cough monitoring terminal 300 of ability
Deng.
Fig. 2 is a kind of block diagram of monitoring server of coughing shown according to an exemplary embodiment.With this hardware configuration
Server can be used for carrying out cough monitoring and be deployed in implementation environment shown in FIG. 1.
It should be noted that the cough monitoring server is the example for adapting to the disclosure, it is not construed as mentioning
Any restrictions to disclosure use scope are supplied.The cough monitoring server can not be construed to need to rely on or necessary
With one or more component in illustrative cough monitoring server 200 shown in Figure 2.
The hardware configuration of the cough monitoring server can generate biggish difference due to the difference of configuration or performance, such as scheme
Shown in 2, cough monitoring server 200 includes: power supply 210, interface 230, at least a memory 250 and an at least centre
Manage device (CPU, Central Processing Units) 270.
Wherein, power supply 210 is used to provide operating voltage for each hardware device in cough monitoring server 200.
Interface 230 includes an at least wired or wireless network interface 231, at least a string and translation interface 233, at least one defeated
Enter output interface 235 and at least USB interface 237 etc., is used for and external device communication.
The carrier that memory 250 is stored as resource, can be read-only memory, random access memory, disk or CD
Deng the resource stored thereon includes operating system 251, application program 253 and data 255 etc., and storage mode can be of short duration
It stores or permanently stores.Wherein, each hardware that operating system 251 is used to manage and control in cough monitoring server 200 is set
Standby and application program 253 can be with realizing calculating and processing of the central processing unit 270 to mass data 255
Windows ServerTM, Mac OS XTM, UnixTM, LinuxTM, FreeBSDTM etc..Application program 253 is based on operation
The computer program that at least one particular job is completed on system 251, may include that an at least module (is not shown in Fig. 2
Out), each module can separately include the series of computation machine readable instruction to cough monitoring server 200.Data 255
Photo, picture, the audio etc. that can be stored in disk.
Central processing unit 270 may include the processor of one or more or more, and be set as through bus and memory
250 communications, for the mass data 255 in operation and processing memory 250.
As described in detail above, the cough monitoring server 200 for being applicable in the disclosure will be read by central processing unit 270
The form of the series of computation machine readable instruction stored in access to memory 250 come complete cough monitoring method.
In the exemplary embodiment, cough monitoring server 200 can be by one or more application specific integrated circuit
At (Application Specific Integrated Circuit, abbreviation ASIC), digital signal processor, digital signal
Manage equipment, programmable logic device, field programmable gate array, controller, microcontroller, microprocessor or other electronic components
It realizes, for executing following methods.Therefore, realize that the present invention is not limited to any specific hardware circuit, software and the two
Combination.
Fig. 3 is a kind of flow chart of cough monitoring method based on deep learning shown according to an exemplary embodiment.
This method can be used for the cough monitoring server 200 in implementation environment shown in Fig. 1, as shown in figure 3, this method, it can be by cough
Monitoring server of coughing 200 executes, and may comprise steps of.
Step S110 pre-processes the audio data of acquisition to obtain several frame tonic trains.
Wherein, audio data can the audio collecting device as shown in Figure 1 in implementation environment collect.It then will acquisition
To audio data upload to cough monitoring server 200 in handled.
In one exemplary embodiment, as shown in Fig. 2, step S110 may further include:
Step S111 carries out Short Time Fourier Transform to the audio data of acquisition and obtains the corresponding sonograph of audio data.
Short Time Fourier Transform (STFT, short-time Fourier transform) is by a Time-Frequency Localization
Window function, it is assumed that audio of coughing in the short time interval that the window function is included is smoothly then Moving Window letter
Number, making audio data is stationary signal in different finite time width, to calculate audio data when each different
The power spectrum at quarter.The power spectrum at each moment is stacked up along time dimension finally, obtains the two dimension letter similar to a width figure
Number form is to get arriving the corresponding sonograph of audio data.
Step S112 carries out the segmentation of sonograph according to specified duration, obtains several frame tonic trains.
Duration is wherein specified to refer to time span corresponding to each frame cough audio frame sequence set by user.It is specified
Duration can set the processing requirement of audio data according to cough monitoring server 200.By specified duration to sound spectrum
Figure is segmented, and if then obtaining dry cough audio frame sequence.
In one embodiment, with audio of coughing for 35s, according to it is specified when it is a length of 4 milliseconds progress sonograph segmentation,
Available 8750 frame cough audio frame sequence, wherein each frame contains 64 frequency components.
Step S130 carries out cough identification to several frame tonic trains, to determine whether audio data is cough audio.
Cough wherein carried out to several audio frame sequences be identified by the mode of deep learning to carry out, i.e., by training after
Neural network model carry out cough identification, thus audio data whether be cough audio.
In one exemplary embodiment, as shown in figure 5, step S130 includes:
Step S131 is extracted from several frame tonic trains and is obtained several local feature vectors of audio data.
Local feature vectors are the features for characterizing tonic train, and audio corresponding to different tonic trains is in time domain
It is different with the feature that is shown on frequency domain, such as tone, loudness, mel-frequency, mel-frequency, mel-frequency cepstrum system
Number (MFCC), linear predictor coefficient (LPC), linear prediction residue error (LPCC) etc..Wherein constructed local feature vectors
It can be the combination based on feature a certain in feature enumerated above or certain several feature to construct.Certainly, the above institute
The feature enumerated is only illustrative examples, can not be considered the limitation to disclosure use scope.
For each frame tonic train one local feature vectors of corresponding building, for characterizing the spy of this frame tonic train
Sign.
Step S132 carries out the full connection of several local feature vectors, obtains the global characteristics vector of audio data.
Step S133 carries out cough Tag Estimation to global characteristics vector, obtains the cough label of audio data.
Whether step S134 is cough audio according to cough label audio data.
Wherein cough label includes the label " cough " for being directed to cough audio and label " the non-cough for non-cough audio
It coughs ".In a specific embodiment, cough label can be indicated by encoding, such as with digital " 1 " expression " cough " label,
" non-cough label " is indicated with digital " 0 ".To be sentenced by coding rule after obtaining the cough label of audio data
Break meaning represented by the cough label, such as cough or non-cough, so according to cough label to judge the audio number
According to whether be cough audio.
When carrying out cough Tag Estimation, prediction obtains the probability P 1 and audio number that audio data is " cough " label respectively
According to the probability P 2 for " non-cough " label, then P1 and P2 are compared, if P1 is greater than P2, export the cough of audio data
Label of coughing is " cough " label；If P1 is less than P2, the cough label for exporting audio data is " non-cough " audio.
In one exemplary embodiment, before carrying out cough identification to audio data, cough is constructed based on neural network
Cough of the identification model for audio data identifies.Figure 10 shows the illustrative diagram of cough identification model, wherein the cough
Identification model of coughing is constructed using Recognition with Recurrent Neural Network, so as to improve cough using the time related information in cough audio
The precision of identification.As shown in Figure 10, which includes: encoder layer, decoder layer, full articulamentum and classification
Layer.
Wherein, encoder layer and decoder layer are used to construct each frame sound by operations such as coding, compression, dimensionality reduction, decodings
The local feature vectors of frequency sequence.Specifically, what encoder layer was formed by 3 layers, including two layers of bidirectional circulating layer and a unidirectional ply,
One of bidirectional circulating layer includes 128 circulation neurons, another bidirectional circulating layer includes 64 circulation neurons, unidirectionally
Layer has 32 circulation neurons.Several frame tonic trains of input pass sequentially through 128 neuronal layers, 64 neuronal layers and 32 minds
Through first layer, subsequently into decoder layer.Several frame tonic trains of input can obtain a tool by 128 circulation neuronal layers
There is the output of 128 dimensions, an output with 64 dimensions can be obtained by recycling neuronal layers by 64, then pass through 32
The layer of a circulation neuron can obtain the output with 32 dimensions.The output for 32 dimensions that encoder layer is exported is made
For the input of decoder layer.
Wherein two-way circulation neuron is to allow neuron when calculating not merely with going through in tonic train
History information can also utilize its following information.The building and transformation for being carried out vector using circulation neuron in encoder layer, are followed
The calculating of ring neuron is and time correlation.
Decoder layer is made of an individual circulation layer, it have 64 long short-term memory (LSTM) neuron, 64
Long memory unit in short-term carries out linear transformation and activation to the output of 32 dimensions of input, and it is right to obtain each frame tonic train institute
The local feature vectors for only one dimension answered.
Further, the LSTM neuron of decoder layer combines attention mechanism, and attention mechanism refers to that decoder exists
When exporting decoding result, an attention range can be also exported, identifying next decoding will pay close attention in sequence
Then which part is exported according to these parts.Because state sometime may be by certain in a time series
The state influence at a little moment is bigger, therefore noted that power mechanism is that the neuron of decoder layer is allowed to calculate shape sometime in decoding
It is average to the implicit variable weighting of different moments when state.It is defeated by being that the LSTM neuron mainly pours into conjunction with attention mechanism
Enter the signal portion of information, and then improves the accuracy of subsequent cough identification.
Full articulamentum has 256 neurons using ReLU activation primitive, the local feature that decoder layer is exported to
Amount is integrated and is converted, that is, carries out the full connection of local feature vectors, obtains the global characteristics vector for the audio data.
Last classification layer uses softmax classifier, carries out classification prediction to the global characteristics vector of input, obtains cough
It coughs label, i.e., " cough " label or " non-cough " label.
Different types of neuron is targetedly set by each layer in neural network, audio sequence can be made full use of
Information in column carries out cough identification, ensure that the precision of prediction of cough disease identification model.
Certainly, cough identification model shown in Fig. 10 is only an illustrative examples, is a preferred cough identification model,
Other neural network models that cough identification may be implemented are equally applicable to the present invention, can not think schematic diagram shown in Fig. 10
It is the limitation to disclosure use scope.
It in one embodiment, further include using sample data to cough before carrying out cough identification using cough identification model
Identification model of coughing is trained.Several sample audio datas are acquired, manually each sample audio data is labeled, for
Sample audio data for audio of coughing is labeled as " coughing ", is labeled as " non-cough for the sample audio data of non-cough audio
It coughs ".Then sample audio data and its corresponding mark are input to the instruction that cough identification model is carried out in cough identification model
Practice.In the training process, for the sample audio data of each input, the classification layer for identification model of coughing one mark of corresponding output
Label, the label of output is compared with the annotation results to the sample audio data, if the label of output with mark not phase
Together, the parameter of cough identification model is adjusted until the label of output is identical as mark；If identical, continue with other samples
Audio data is trained.Finally until the accuracy of identification of cough identification model reaches permissible accuracy, then completing to cough identifies
The training of model.The cough identification model that training is completed is used for the cough identification of audio data.
And step S140, Application on Voiceprint Recognition is carried out to several frame tonic trains, to determine the sounder of audio data.
Application on Voiceprint Recognition is that the identification of sounder identity is carried out by the characteristic voice of sounder, i.e., according to an audio data
Audio features carry out sounder identification.
In one embodiment, in one exemplary embodiment, as shown in fig. 6, step S140 includes:
Step S141 is extracted from several frame tonic trains and is obtained the vocal print feature of audio data.
Step S142 carries out sounder Tag Estimation according to vocal print feature, obtains the sounder label of audio data.
Step S143 determines the sounder of audio data according to sounder label.
Wherein Application on Voiceprint Recognition can be carried out by Application on Voiceprint Recognition model, Application on Voiceprint Recognition model is also based on neural network and is built
It is vertical.Application on Voiceprint Recognition model carries out the process of Application on Voiceprint Recognition as shown in step S141, S142, S143.
Extracted vocal print feature can be the tamber characteristic of audio data, it is, of course, also possible to be that other can characterize sound
Frequency is according to the feature for corresponding to sounder feature, such as tamber characteristic combination mel-frequency, mel-frequency cepstrum coefficient, linear prediction
Some or certain several features in cepstrum coefficient.Then sounder label is carried out according to extracted vocal print feature.Finally
The sounder of audio data is determined according to sounder label.
Before carrying out Application on Voiceprint Recognition using Application on Voiceprint Recognition model, need to be trained Application on Voiceprint Recognition model, i.e., with several
The audio data of people is trained, and marks corresponding sounder to each audio data, passes through the hair of audio data and mark
The training of sound person progress Application on Voiceprint Recognition model, so that the Application on Voiceprint Recognition model after training can be determined for the audio of an input and is somebody's turn to do
The sounder of audio.
Step S150, if audio data is cough audio, according to audio data to the cough monitoring data of sounder
It is updated processing.
If audio data is cough audio, Cough length, cough number etc. in the audio data can be calculated
Then cough information will be added in the cough monitoring data monitored from cough information obtained in the audio data, real
The update for monitoring data of now coughing.
In one exemplary embodiment, as shown in fig. 7, step S150 includes:
Step S151 is extracted from audio data if audio data is cough audio and is obtained timestamp information.
Audio generation time corresponding to any node audio in timestamp information, that is, audio data.Wherein timestamp information
It can be the timestamp information that audio data is automatically recorded when audio collecting device 100 acquires audio data and by the timestamp
Information is added in audio data, thus in step S151 can from audio data extraction time stab information.
Step S152 generates Cough length information and the cough frequency according to cough audio and corresponding timestamp information
Information.
The multiple cough that may include sounder in audio of wherein coughing passes through analysis cough audio and cough audio pair
The timestamp information answered, the available Cough length information for the cough audio and cough frequency information.
Wherein Cough length information includes but is not limited to: every time at the beginning of cough, end time for coughing every time, every
The duration of secondary cough.Frequency information of coughing includes but is not limited to the cough in the corresponding time range of this section cough audio
Number, cough frequency etc..
In one exemplary embodiment, as shown in figure 8, step S152 includes:
Step S210 carries out end-point detection to cough audio to determine the beginning and end coughed every time in cough audio.
End-point detection is exactly that effective audio section is detected from continuous audio stream.In the scheme of the disclosure, cough
Corresponding audio is effective audio end.Breaking point detection includes two aspects, detects the starting point of effective audio, detects
The terminal of effective audio.Breaking point detection determines the beginning and end of effective audio by the energy value of audio.It coughed from non-
It crosses to cough, the reflection on audio power is: audio power sharply increases.And it is transitioned into non-cough from cough, in audio energy
Reflection in amount is that audio power drastically reduces.
So can by set energy threshold, if in a certain node, the energy of continuous N frame tonic train before the node
Value is greater than the energy of setting in the energy value of the next continuous N frame tonic train of the node lower than the energy threshold set
Threshold value, then the node can be determined as the starting point once coughed.Likewise, if in a certain node, continuous N frame before the node
The energy value of tonic train is greater than the energy threshold of setting, and the energy value of the next continuous N frame tonic train of the node is small
In the energy threshold of setting, then the node can be determined as the terminal once coughed.
In the primary cough of corresponding sounder, this time audio caused by cough is continuous, can by end-point detection
To determine the beginning and end for corresponding audio of this time coughing.Thus in audio data collected, it may be possible to corresponding one
The audio of secondary cough, it is also possible to the audio repeatedly coughed.When being the audio repeatedly coughed for audio data collected, often
Audio corresponding to secondary cough is effective audio of end-point detection, and the audio between cough in the period may be background twice
Noise is mute, and the audio between cough in the period is invalid audio twice.
Step S230, the timestamp information according to corresponding to the beginning and end coughed every time obtain Cough length information, with
And the points or terminal number coughed in statistics cough audio obtain cough frequency information.
Cough time started, cough end time and the cough that wherein Cough length information is including but not limited to coughed every time
Duration.Cough frequency information includes but is not limited to correspond to the number coughed in the period, cough frequency in the cough audio.
When the cough that wherein the secondary cough can be obtained in the timestamp information according to corresponding to the starting point coughed every time starts
Between, the cough end time of the secondary cough can be obtained according to timestamp information corresponding to the terminal coughed every time.Then lead to
It spends the cough time started and the cough duration of the secondary cough is calculated in the cough end time.
Certainly in cough audio has been determined after the beginning and end coughed every time, can by count starting point number or
The number of person's terminal obtains number of coughing in this section cough audio, and the frequency of cough can be obtained by obtained cough number
Rate.
Step S153, the cough in short-term using Cough length information and cough frequency information as sounder in audio data
Monitoring data.
Step S154, the monitoring data that will cough in short-term upload in monitor database, to update sounding in monitor database
The cough monitoring data of person.
Monitor database be used for store sounder audio data and corresponding cough monitoring data, wherein each send out
Sound person corresponds to a subdata base, and the cough monitoring data of the sounder can be obtained from the subdata base.For example, in hospital
Application scenarios in, monitor database can be the medical data base for patient, so as to from the subdata of the patient
Transfer the cough monitoring data of the patient.In a particular embodiment, in the subdata base of patient can also comprising patient case,
The information such as treatment record.
Obtained cough monitoring data be based on before the audio data collected data analyze, so
The monitoring data of cough in short-term obtained based on the audio data analysis are added in obtained cough monitoring data, are carried out
The update for monitoring data of coughing, the including but not limited to update of Cough length information, the update for frequency information of coughing.
For monitoring data of coughing after monitor database update, related personnel can view cough by the monitor database
Monitoring data are realized to the remote real-time monitoring of patient's cough monitoring situation, are not needed special messenger and supervise near sounder
Control.In a particular embodiment, such as in hospital, sounder is patient, then the attending physician of the patient or nurse can be at any time
Cough monitoring data are viewed in monitor database, further, doctor can also be supervised by the cough within a period of time
Measured data carries out the prediction of patient's state of an illness and recovery situation.
After monitoring data of coughing update, updated cough monitoring data can be carried out by cough monitoring terminal 300
It has been shown that, cough monitoring terminal 300 can be the ward positioned at the house where sounder, such as patient, thus either family
Category, nurse, doctor can monitor the cough monitoring situation that terminal 300 quickly understands patient by the cough.
Shown by a particular embodiment, being adjusted by setting time range in cough monitoring terminal 300
Cough monitoring data.For example, if the cough frequency of statistics is the cough frequency in 1 hour, it is only whole in cough monitoring
The cough frequency in 1 hour is shown on end 300, and carries out real-time update.Cough monitoring data in certain other times can
To be stored in specified data library, it can transfer and check in real time for user.
In one exemplary embodiment, the method for the embodiment further include:
Step S170 judges whether the cough frequency of sounder exceeds the safe frequency according to cough frequency information.
Step S190 generates information warning if exceeding the safe frequency.
The cough frequency may cause expiratory dyspnea beyond the safe frequency, thus by setting the safe frequency, if beyond peace
Full range, then generate information warning.In a particular embodiment, external warning device (example can be triggered by the information warning
Such as buzzer, alarm lamp) movement, so as to allow the personnel of surrounding to learn and handle in time.Certainly, warning generated
Information can also be broadcasted automatically by voice, and related personnel is allowed to recognize the situation in time.
It is worth noting that the technical solution of the disclosure cannot only be applied in the scene of hospital, for needing to monitor
The patient of status of cough can be used for the scenes such as home for destitute, sanatorium, family and carry out cough monitoring, herein without specific
It limits.
Following is embodiment of the present disclosure, can be used for executing that the above-mentioned mobile terminal 110 of the disclosure executes based on depth
Spend the cough monitoring method embodiment of study.For those undisclosed details in the apparatus embodiments, disclosure base is please referred to
In the cough monitoring method embodiment of deep learning.
Figure 11 is a kind of block diagram of cough monitoring device based on deep learning shown according to an exemplary embodiment, should
Device can be used in the cough monitoring server 200 of implementation environment shown in Fig. 1, execute shown in any of the above embodiment of the method
The all or part of step of cough monitoring method based on deep learning.As shown in figure 11, which includes but unlimited
In: preprocessing module 110, cough identification module 130, voiceprint identification module 140 and monitoring data update module 150.
Preprocessing module 110 is configured as executing: being pre-processed to obtain several frame audio sequences to the audio data of acquisition
Column.
Cough identification module 130, the module connect with preprocessing module 110, are configured as executing: to several frame audio sequences
Column carry out cough identification, to determine whether audio data is cough audio.And
Voiceprint identification module 140, the module are connect with preprocessing module 110, are configured as executing: to several frame audio sequences
Column carry out Application on Voiceprint Recognition, to determine the sounder of audio data.
Monitoring data update module 150, the module connect with cough identification module 130, voiceprint identification module 140, are matched
It is set to execution: if audio data is cough audio, being updated according to cough monitoring data of the audio data to sounder
Processing.
In one embodiment, preprocessing module 110 includes:
Short Time Fourier Transform unit is configured as executing: carrying out Short Time Fourier Transform to the audio data of acquisition and obtains
To the corresponding sonograph of audio data.
Segmenting unit is configured as executing: carrying out the segmentation of sonograph according to specified duration, obtains several frame audio sequences
Column.
In one embodiment, cough identification module includes:
Local feature vectors extraction unit is configured as executing: extracting from several frame tonic trains and obtains audio data
Several local feature vectors.
Full connection unit is configured as executing: carrying out the full connection of several local feature vectors, obtains the complete of audio data
Office's feature vector.
Cough label prediction unit, is configured as executing: carrying out cough Tag Estimation to global characteristics vector, obtain audio
The cough label of data.
Whether interpretation unit is configured as executing: being cough audio according to cough label audio data.
In one embodiment, voiceprint identification module includes:
Vocal print feature extraction unit is configured as executing: extracting from several frame tonic trains and obtains the sound of audio data
Line feature.
Sounder Tag Estimation module, is configured as executing: carrying out sounder Tag Estimation according to vocal print feature, obtains sound
The sounder label of frequency evidence.
Sounder determining module is configured as executing: the sounder of audio data is determined according to sounder label.
In one embodiment, monitoring data update module includes:
Timestamp information extraction unit is configured as executing: if audio data is cough audio, from audio data
Extraction obtains timestamp information.
Cough length information and cough frequency information generating unit, are configured as executing: right according to cough audio and institute
The timestamp information answered generates Cough length information and cough frequency information.
Cough monitoring data acquiring unit in short-term, is configured as executing: Cough length information and cough frequency information are made
The cough monitoring data in short-term for being sounder in audio data.
Monitoring data updating unit is configured as executing: the monitoring data that will cough in short-term upload in monitor database, with
Update the cough monitoring data of sounder in monitor database.
In one embodiment, Cough length information and cough frequency information generating unit include:
End-point detection unit is configured as executing: carrying out end-point detection to cough audio to determine in cough audio every time
The beginning and end of cough.
Acquiring unit, be configured as executing: the timestamp information according to corresponding to the beginning and end coughed every time obtains cough
The points or terminal number coughed in the temporal information, and statistics cough audio of coughing obtain cough frequency information.
Modules/unit function and the realization process of effect are specifically detailed in above-mentioned based on deep learning in above-mentioned apparatus
Cough monitoring method in correspond to the realization process of step, details are not described herein.
It is appreciated that these module/units can by hardware, software, or a combination of both realize.When in hardware
When realization, these modules may be embodied as one or more hardware modules, such as one or more specific integrated circuits.When with soft
When part mode is realized, these modules may be embodied as the one or more computer journeys executed on the one or more processors
Sequence, such as the program being stored in performed by the central processing unit 270 of Fig. 2 in memory 250.
Optionally, the disclosure also provides a kind of cough monitoring device 1000 based on deep learning, as shown in figure 12, packet
It includes:
Processor 1001；And
Memory 1002 for 1001 executable instruction of storage processor；
Wherein, executable instruction by processor 1001 execute when realize above method embodiment in it is any shown in all or
Person's part steps, executable instruction can be computer-readable instruction, in processor work, can pass through data line/communication
Line 1003 reads the computer-readable instruction in memory 1002 and execution.
The concrete mode that the processor of device in the embodiment executes operation is somebody's turn to do related based on deep learning
Detailed description is performed in the embodiment for monitoring method of coughing, no detailed explanation will be given here.
Optionally, the disclosure also provides a kind of computer readable storage medium, is stored thereon with computer program, computer
The cough monitoring method based on deep learning shown in any of the above embodiment is realized when program is executed by processor.Wherein, should
Computer readable storage medium can be the provisional and non-transitorycomputer readable storage medium for including computer program, should
Memory 250 of the computer readable storage medium for example including computer program, above-mentioned computer program can be by cough monitoring clothes
The central processing unit 270 of business device 200 is executed to complete the above-mentioned cough monitoring method based on deep learning.
It should be understood that the present invention is not limited to the precise structure already described above and shown in the accompanying drawings, and
And various modifications and change can executed without departing from the scope.The scope of the present invention is limited only by the attached claims.
Claims (10)
1. a kind of cough monitoring method based on deep learning characterized by comprising
The audio data of acquisition is pre-processed to obtain several frame tonic trains；
Cough identification is carried out to several frame tonic trains, whether is cough audio with the determination audio data；And
Application on Voiceprint Recognition is carried out to several frame tonic trains, with the sounder of the determination audio data；
If the audio data is cough audio, according to the audio data to the cough monitoring data of the sounder into
Row update processing.
2. if the method according to claim 1, wherein the audio data of described pair of acquisition is pre-processed to obtain
Dry frame tonic train, comprising:
Short Time Fourier Transform is carried out to the audio data of acquisition and obtains the corresponding sonograph of the audio data；
The segmentation that the sonograph is carried out according to specified duration obtains several frame tonic trains.
3. the method according to claim 1, wherein described carry out cough identification to several frame tonic trains
It whether is cough audio with the determination audio data, comprising:
It is extracted from several frame tonic trains and obtains several local feature vectors of the audio data；
The full connection for carrying out several local feature vectors obtains the global characteristics vector of the audio data；
Cough Tag Estimation is carried out to the global characteristics vector, obtains the cough label of the audio data；
Judge whether the audio data is cough audio according to the cough label.
4. the method according to claim 1, wherein described carry out vocal print knowledge to several frame tonic trains
Not, with the sounder of the determination audio data, comprising:
It is extracted from several frame tonic trains and obtains the vocal print feature of the audio data；
Sounder Tag Estimation is carried out according to the vocal print feature, obtains the sounder label of the audio data；
The sounder of the audio data is determined according to the sounder label.
5. method according to any one of claims 1 to 4, which is characterized in that if the audio data is cough
Audio is then updated processing according to cough monitoring data of the audio data to the sounder, comprising:
If the audio data is cough audio, is extracted from the audio data and obtain timestamp information；
Cough length information and cough frequency information are generated according to the cough audio and the corresponding timestamp information；
Cough in short-term using the Cough length information and cough frequency information as the sounder in the audio data
Monitoring data；
The monitoring data of cough in short-term are uploaded in monitor database, to update sounder described in the monitor database
Cough monitoring data.
6. according to the method described in claim 5, it is characterized in that, it is described according to the cough audio and it is corresponding described in
Timestamp information generates Cough length information and cough frequency information, comprising:
The beginning and end that end-point detection is carried out to cough every time in the determination cough audio to the cough audio；
The Cough length information, and statistics are obtained according to timestamp information corresponding to the beginning and end coughed every time
The points or terminal number coughed in the cough audio obtain the cough frequency information.
7. according to the method described in claim 5, it is characterized in that, the method also includes:
Judge whether the cough frequency of the sounder exceeds the safe frequency according to the cough frequency information；
If exceeding the safe frequency, information warning is generated.
8. a kind of cough monitoring device based on deep learning characterized by comprising
Preprocessing module is configured as executing: being pre-processed to obtain several frame tonic trains to the audio data of acquisition；
Cough identification module, is configured as executing: cough identification is carried out to several frame tonic trains, with the determination audio
Whether data are cough audio；And
Voiceprint identification module is configured as executing: Application on Voiceprint Recognition is carried out to several frame tonic trains, with the determination audio
The sounder of data；
Monitoring data update module is configured as executing: if the audio data is cough audio, according to the audio number
Processing is updated according to the cough monitoring data to the sounder.
9. a kind of cough monitoring device based on deep learning characterized by comprising
Processor；And
Memory for storage processor executable instruction；
Wherein, method as described in any one of claim 1 to 7 is realized when the executable instruction is executed by the processor.
10. a kind of computer readable storage medium, is stored thereon with computer program, which is characterized in that the computer program
The method as described in any one of claims 1 to 7 is realized when being executed by processor.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811259953.4A CN109360584A (en) | 2018-10-26 | 2018-10-26 | Cough monitoring method and device based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811259953.4A CN109360584A (en) | 2018-10-26 | 2018-10-26 | Cough monitoring method and device based on deep learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109360584A true CN109360584A (en) | 2019-02-19 |
Family
ID=65346788
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811259953.4A Pending CN109360584A (en) | 2018-10-26 | 2018-10-26 | Cough monitoring method and device based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109360584A (en) |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110111815A (en) * | 2019-04-16 | 2019-08-09 | 平安科技（深圳）有限公司 | Animal anomaly sound monitoring method and device, storage medium, electronic equipment |
| CN110890102A (en) * | 2019-09-07 | 2020-03-17 | 创新奇智(重庆)科技有限公司 | Engine defect detection algorithm based on RNN voiceprint recognition |
| CN111028862A (en) * | 2019-12-17 | 2020-04-17 | 北京海益同展信息科技有限公司 | Method, apparatus, computer device and storage medium for processing voice data |
| CN111091849A (en) * | 2020-03-03 | 2020-05-01 | 龙马智芯（珠海横琴）科技有限公司 | Snore identification method and device, storage medium snore stopping equipment and processor |
| CN111524537A (en) * | 2020-03-24 | 2020-08-11 | 苏州数言信息技术有限公司 | Method for identifying coughing and sneezing aiming at real-time voice flow |
| CN111860415A (en) * | 2020-07-29 | 2020-10-30 | 浙江大华技术股份有限公司 | Cough detection method, device, storage medium and electronic device |
| CN112687290A (en) * | 2020-12-30 | 2021-04-20 | 同济大学 | Compressed cough automatic detection method and embedded device |
| CN113327592A (en) * | 2021-04-20 | 2021-08-31 | 青岛海尔空调器有限总公司 | Air conditioning system control method and device, electronic equipment and storage medium |
| WO2021169742A1 (en) * | 2020-02-26 | 2021-09-02 | Oppo广东移动通信有限公司 | Method and device for predicting operating state of transportation means, and terminal and storage medium |
| CN113488071A (en) * | 2021-07-16 | 2021-10-08 | 河南牧原智能科技有限公司 | Pig cough recognition method, device, equipment and readable storage medium |
| CN113948109A (en) * | 2021-10-14 | 2022-01-18 | 广州蓝仕威克软件开发有限公司 | System for recognizing physiological phenomenon based on voice |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN202235373U (en) * | 2011-09-05 | 2012-05-30 | 苏州大学 | Cough monitor |
| CN202723829U (en) * | 2012-04-06 | 2013-02-13 | 肖遥 | Intelligent cough monitoring and evaluation system |
| CN107242857A (en) * | 2017-06-12 | 2017-10-13 | 南开大学 | The intelligent traditional Chinese medical science based on deep learning integrates diagnosis and therapy system |
| CN108403119A (en) * | 2018-02-26 | 2018-08-17 | 深圳乐测物联网科技有限公司 | A kind of cough detection method |
| CN108701469A (en) * | 2017-07-31 | 2018-10-23 | 深圳和而泰智能家居科技有限公司 | Cough sound recognition methods, equipment and storage medium |
-
2018
- 2018-10-26 CN CN201811259953.4A patent/CN109360584A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN202235373U (en) * | 2011-09-05 | 2012-05-30 | 苏州大学 | Cough monitor |
| CN202723829U (en) * | 2012-04-06 | 2013-02-13 | 肖遥 | Intelligent cough monitoring and evaluation system |
| CN107242857A (en) * | 2017-06-12 | 2017-10-13 | 南开大学 | The intelligent traditional Chinese medical science based on deep learning integrates diagnosis and therapy system |
| CN108701469A (en) * | 2017-07-31 | 2018-10-23 | 深圳和而泰智能家居科技有限公司 | Cough sound recognition methods, equipment and storage medium |
| CN108403119A (en) * | 2018-02-26 | 2018-08-17 | 深圳乐测物联网科技有限公司 | A kind of cough detection method |
Non-Patent Citations (1)
| Title |
|---|
| JUSTICE AMOH 等: "Deep Neural Networks for Identifying Cough Sounds", 《IEEE TRANSACTIONS ON BIOMEDICAL CIRCUITS AND SYSTEMS》 * |
Cited By (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110111815A (en) * | 2019-04-16 | 2019-08-09 | 平安科技（深圳）有限公司 | Animal anomaly sound monitoring method and device, storage medium, electronic equipment |
| CN110890102A (en) * | 2019-09-07 | 2020-03-17 | 创新奇智(重庆)科技有限公司 | Engine defect detection algorithm based on RNN voiceprint recognition |
| CN111028862B (en) * | 2019-12-17 | 2022-09-06 | 京东科技信息技术有限公司 | Method, apparatus, computer device and storage medium for processing voice data |
| CN111028862A (en) * | 2019-12-17 | 2020-04-17 | 北京海益同展信息科技有限公司 | Method, apparatus, computer device and storage medium for processing voice data |
| WO2021169742A1 (en) * | 2020-02-26 | 2021-09-02 | Oppo广东移动通信有限公司 | Method and device for predicting operating state of transportation means, and terminal and storage medium |
| CN111091849A (en) * | 2020-03-03 | 2020-05-01 | 龙马智芯（珠海横琴）科技有限公司 | Snore identification method and device, storage medium snore stopping equipment and processor |
| CN111524537A (en) * | 2020-03-24 | 2020-08-11 | 苏州数言信息技术有限公司 | Method for identifying coughing and sneezing aiming at real-time voice flow |
| CN111524537B (en) * | 2020-03-24 | 2023-04-14 | 苏州数言信息技术有限公司 | Cough and sneeze identification method aiming at real-time voice flow |
| CN111860415A (en) * | 2020-07-29 | 2020-10-30 | 浙江大华技术股份有限公司 | Cough detection method, device, storage medium and electronic device |
| CN112687290A (en) * | 2020-12-30 | 2021-04-20 | 同济大学 | Compressed cough automatic detection method and embedded device |
| CN112687290B (en) * | 2020-12-30 | 2022-09-20 | 同济大学 | Compressed cough automatic detection method and embedded device |
| CN113327592B (en) * | 2021-04-20 | 2022-08-19 | 青岛海尔空调器有限总公司 | Air conditioning system control method and device, electronic equipment and storage medium |
| CN113327592A (en) * | 2021-04-20 | 2021-08-31 | 青岛海尔空调器有限总公司 | Air conditioning system control method and device, electronic equipment and storage medium |
| CN113488071A (en) * | 2021-07-16 | 2021-10-08 | 河南牧原智能科技有限公司 | Pig cough recognition method, device, equipment and readable storage medium |
| CN113948109A (en) * | 2021-10-14 | 2022-01-18 | 广州蓝仕威克软件开发有限公司 | System for recognizing physiological phenomenon based on voice |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109360584A (en) | Cough monitoring method and device based on deep learning | |
| CN109431507A (en) | Cough disease identification method and device based on deep learning | |
| AU2013239327B2 (en) | A method and apparatus for processing patient sounds | |
| Istrate et al. | Information extraction from sound for medical telemonitoring | |
| Vacher et al. | The sweet-home project: Audio technology in smart homes to improve well-being and reliance | |
| Aminikhanghahi et al. | Using change point detection to automate daily activity segmentation | |
| US20240127078A1 (en) | Transferring learning in classifier-based sensing systems | |
| CN109044396B (en) | Intelligent heart sound identification method based on bidirectional long-time and short-time memory neural network | |
| US20210267488A1 (en) | Sensing system and method for monitoring time-dependent processes | |
| CN111161715A (en) | Specific sound event retrieval and positioning method based on sequence classification | |
| JP2015207179A (en) | behavior understanding system and program | |
| Ghods et al. | Activity2vec: Learning adl embeddings from sensor data with a sequence-to-sequence model | |
| CN110111815A (en) | Animal anomaly sound monitoring method and device, storage medium, electronic equipment | |
| CN111429943B (en) | Joint detection method for music and relative loudness of music in audio | |
| CN113080907B (en) | Pulse wave signal processing method and device | |
| Min et al. | Semi-supervised event detection using higher order statistics for multidimensional time series accelerometer data | |
| CN114830257A (en) | Cough detection system and method | |
| CN117198468A (en) | Intervention scheme intelligent management system based on behavior recognition and data analysis | |
| CN114581749B (en) | Audio-visual feature fusion target behavior identification method and device and application | |
| CN115205972A (en) | Six-step hand washing evaluation method and device based on deep learning | |
| Pettinati et al. | Automatic and robust identification of spontaneous coughs from covid-19 patients | |
| Hervás et al. | homeSound: A High Performance Platform for Massive Data Acquisition and Processing in Ambient Assisted Living Environments. | |
| Wang et al. | Epileptic Seizures Prediction Based on Unsupervised Learning for Feature Extraction | |
| Kapoor et al. | IoT based real-time face detection and recognition system | |
| CN116052725B (en) | Fine granularity borborygmus recognition method and device based on deep neural network |
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: 20190219 |
|
| RJ01 | Rejection of invention patent application after publication |