CN115762551A - Snore detection method and device, computer equipment and storage medium - Google Patents
Snore detection method and device, computer equipment and storage medium Download PDFInfo
- Publication number
- CN115762551A CN115762551A CN202211457181.1A CN202211457181A CN115762551A CN 115762551 A CN115762551 A CN 115762551A CN 202211457181 A CN202211457181 A CN 202211457181A CN 115762551 A CN115762551 A CN 115762551A
- Authority
- CN
- China
- Prior art keywords
- snore
- snore signal
- signal
- segment
- signal data
- 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
- 206010041235 Snoring Diseases 0.000 title claims abstract description 499
- 238000001514 detection method Methods 0.000 title claims abstract description 49
- 238000000034 method Methods 0.000 claims abstract description 73
- 230000011218 segmentation Effects 0.000 claims abstract description 43
- 238000012545 processing Methods 0.000 claims abstract description 39
- 238000000354 decomposition reaction Methods 0.000 claims abstract description 29
- 238000003062 neural network model Methods 0.000 claims abstract description 23
- 230000009467 reduction Effects 0.000 claims abstract description 19
- 238000000605 extraction Methods 0.000 claims abstract description 18
- 239000012634 fragment Substances 0.000 claims description 39
- 238000005070 sampling Methods 0.000 claims description 27
- 238000004590 computer program Methods 0.000 claims description 13
- 238000012216 screening Methods 0.000 claims description 11
- 238000001914 filtration Methods 0.000 claims description 10
- 238000007781 pre-processing Methods 0.000 claims description 8
- 230000005236 sound signal Effects 0.000 description 12
- 230000008569 process Effects 0.000 description 11
- 230000006870 function Effects 0.000 description 6
- 238000005457 optimization Methods 0.000 description 5
- 238000010586 diagram Methods 0.000 description 4
- 238000004458 analytical method Methods 0.000 description 3
- 238000011176 pooling Methods 0.000 description 3
- 230000004913 activation Effects 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 230000036541 health Effects 0.000 description 2
- 230000000670 limiting effect Effects 0.000 description 2
- 208000001797 obstructive sleep apnea Diseases 0.000 description 2
- 230000036961 partial effect Effects 0.000 description 2
- 230000002829 reductive effect Effects 0.000 description 2
- 238000001228 spectrum Methods 0.000 description 2
- 238000013526 transfer learning Methods 0.000 description 2
- 208000024172 Cardiovascular disease Diseases 0.000 description 1
- 208000008784 apnea Diseases 0.000 description 1
- 238000004422 calculation algorithm Methods 0.000 description 1
- 208000026106 cerebrovascular disease Diseases 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 238000010219 correlation analysis Methods 0.000 description 1
- 230000002526 effect on cardiovascular system Effects 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 230000037433 frameshift Effects 0.000 description 1
- SYHGEUNFJIGTRX-UHFFFAOYSA-N methylenedioxypyrovalerone Chemical compound C=1C=C2OCOC2=CC=1C(=O)C(CCC)N1CCCC1 SYHGEUNFJIGTRX-UHFFFAOYSA-N 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 230000008447 perception Effects 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 230000029058 respiratory gaseous exchange Effects 0.000 description 1
- 230000000717 retained effect Effects 0.000 description 1
- 238000000926 separation method Methods 0.000 description 1
- 230000003595 spectral effect Effects 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 208000024891 symptom Diseases 0.000 description 1
- 230000001360 synchronised effect Effects 0.000 description 1
- 208000011580 syndromic disease Diseases 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Images
Landscapes
- Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)
Abstract
The invention discloses a snore detecting method, a device, computer equipment and a storage medium, which are used for acquiring initial snore signal data, and performing noise reduction processing on the initial snore signal data by adopting a variational modal decomposition method to generate reconstructed snore signal data; adopting a voice endpoint detection method to segment the reconstructed snore signal data to generate a snore signal segmentation sequence; performing feature extraction on the snore signal segmentation sequence to generate snore signal features; inputting the snore signal characteristics into a preset neural network model, and outputting target signal characteristics; constructing a full-connection layer to classify the target signal characteristics and output a snore detection result; therefore, the accuracy of the detection result output when the snore signal is detected is improved.
Description
Technical Field
The invention relates to the technical field of sound signal detection, in particular to a snore detecting method, a snore detecting device, computer equipment and a storage medium.
Background
Snoring is one of the important symptoms of sleep disordered breathing and is closely related to Obstructive Sleep Apnea (OSA) syndrome. Severe snoring can even induce cardiovascular and cerebrovascular diseases caused by apnea, which poses a significant threat to life health. The accurate detection of the snore event has important significance for sleep health monitoring.
The traditional snore detecting method generally utilizes a signal processing algorithm to directly segment original snore signals and extract characteristics such as time-frequency domain of signal fragments, and then carries out snore fragment classification based on the artificial characteristics. However, the original snore signal collected in the actual environment often contains a lot of environmental noises, such as human breath, automobile whistling outside the window, wind noise, etc., and the traditional method often ignores or is difficult to effectively implement noise reduction processing on the original signal before the original signal is segmented; in addition, in terms of original signal segmentation, the conventional method easily ignores the condition that some segments of the obtained sound segments overlap, thereby affecting the accuracy of snore detection.
Disclosure of Invention
The embodiment of the invention provides a snore detecting method, a snore detecting device, computer equipment and a storage medium, and aims to solve the problem that the accuracy of snore detection is not high.
A snore detecting method, comprising:
acquiring initial snore signal data, and performing noise reduction processing on the initial snore signal data by adopting a variational modal decomposition method to generate reconstructed snore signal data;
adopting a voice end point detection method to carry out segmentation processing on the reconstructed snore signal data to generate a snore signal segmentation sequence;
performing feature extraction on the snore signal segmentation sequence to generate snore signal features;
inputting the snore signal characteristics into a preset neural network model, and outputting target signal characteristics;
and constructing a full connection layer to classify the target signal characteristics and outputting snore detection results.
A snore detecting device comprising:
the device comprises a first acquisition module, a second acquisition module and a third acquisition module, wherein the first acquisition module is used for acquiring initial snore signal data and performing noise reduction processing on the initial snore signal data by adopting a variational modal decomposition method to generate reconstructed snore signal data;
the segmentation processing module is used for carrying out segmentation processing on the reconstructed snore signal data by adopting a voice endpoint detection method to generate a snore signal segmentation sequence;
the characteristic extraction module is used for carrying out characteristic extraction on the snore signal segmented sequence to generate snore signal characteristics;
the detection module is used for inputting the snore signal characteristics into a preset neural network model and outputting target signal characteristics;
and the classification module is used for constructing a full connection layer to classify the target signal characteristics and outputting a snore detection result.
A computer device comprising a memory, a processor and a computer program stored in said memory and executable on said processor, said processor implementing the above snore detection method when executing said computer program.
A computer-readable storage medium, in which a computer program is stored which, when being executed by a processor, carries out the snore detecting method as described above.
The snore detecting method, the snore detecting device, the computer equipment and the storage medium acquire initial snore signal data, and perform noise reduction processing on the initial snore signal data by adopting a variational modal decomposition method to generate reconstructed snore signal data; adopting a voice endpoint detection method to segment the reconstructed snore signal data to generate a snore signal segmentation sequence; performing feature extraction on the snore signal segmentation sequence to generate snore signal features; inputting the snore signal characteristics into a preset neural network model, and outputting target signal characteristics; constructing a full-connection layer to classify the target signal characteristics and output a snore detection result; in the embodiment, the whole-process snore detection of the snore signal is realized by performing variation modal decomposition noise reduction on the obtained initial snore signal data, segmenting the reconstructed snore signal data by using a voice endpoint detection method, and extracting and combining a feature extraction and transfer learning method; and further, the accuracy of the detection result output when the snore signal is detected is improved.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings needed to be used in the description of the embodiments of the present invention will be briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art that other drawings can be obtained according to these drawings without inventive labor.
FIG. 1 is a schematic diagram of an application environment of the snore detecting method in an embodiment of the present invention;
FIG. 2 is a flow chart of a snore detecting method according to an embodiment of the invention;
FIG. 3 is another flow chart of a snore detecting method according to an embodiment of the present invention;
FIG. 4 is another flow chart of a snore detecting method according to an embodiment of the present invention;
FIG. 5 is another flow chart of a snore detecting method according to an embodiment of the present invention;
FIG. 6 is another flow chart of a snore detecting method according to an embodiment of the present invention;
FIG. 7 is a flow chart of the snore detecting device in an embodiment of the present invention;
FIG. 8 is another flow chart of the snore detecting device in an embodiment of the present invention;
FIG. 9 is a schematic diagram of a computer device according to an embodiment of the invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, not all, embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The snore detecting method provided by the embodiment of the invention can be applied to an application environment shown in figure 1. Specifically, the snore detecting method is applied to a snore detecting system, the snore detecting system comprises a client and a server shown in fig. 1, and the client and the server are communicated through a network and used for solving the problem that the snore detecting accuracy is not high. The client is also called a client, and refers to a program corresponding to the server and providing local services to the client. The client may be installed on, but is not limited to, various personal computers, laptops, smartphones, tablets, and portable wearable devices. The server can be implemented by an independent server or a server cluster composed of a plurality of servers.
In an embodiment, as shown in fig. 2, a snore detecting method is provided, which is described by taking the method as an example of being applied to the server side in fig. 1, and includes the following steps:
s10: obtaining initial snore signal data, and carrying out noise reduction processing on the initial snore signal data by adopting a variational modal decomposition method to generate reconstructed snore signal data.
Wherein, the initial snore signal data is a signal which is obtained in advance and is to be subjected to noise reduction processing. In this embodiment, the initial snore signal data is the snore signal data after the interference has been removed by preprocessing.
The variational modal decomposition method is a self-adaptive and completely non-recursive modal variational and signal processing method. In signal processing, the metamorphic mode decomposition is a signal decomposition estimation method. According to the method, in the process of acquiring the decomposition components, the frequency center and the bandwidth of each component are determined by iteratively searching the optimal solution of the variation model, so that the frequency domain subdivision of the signal and the effective separation of each component can be adaptively realized.
Because the variational modal decomposition method has the characteristic of self-determining the number of modal decompositions, the time sequence non-stationarity with high complexity and strong nonlinearity can be reduced. Therefore, in this embodiment, noise reduction processing is performed on the initial snore signal data by using a variational modal decomposition method, and a specific implementation manner is that firstly, variational modal decomposition is performed on the preprocessed initial snore signal data to generate K modal components, where K is a positive integer, then, the energy entropy of each modal component is calculated, and then, the corresponding ith modal component of the local minimum value is determined from the energy entropy of each modal component, where i is a positive integer smaller than K; and finally, filtering the 1 st to ith modal components, and reconstructing the (i + 1) th to Kth modal components to obtain variation modal decomposition to generate reconstructed snore signal data.
S20: and adopting a voice end point detection method to carry out segmentation processing on the reconstructed snore signal data to generate a snore signal segmentation sequence.
The End-point detection (EPD) is to find the beginning and ending positions of an audio signal, and may be called as spech Detect i on or VAD (Vo i ce Act i ty Detect i on). Endpoint detection in speech processing and recognition may be done on a time domain basis, or on a frequency domain basis. In this embodiment, a speech endpoint detection method is adopted to segment the reconstructed snore signal data to generate a snore signal segmentation sequence. Wherein. The sequence of segments of the snore signal is a sequence consisting of several small segments/frames of the signal, each segment being called a frame. And (3) carrying out segmentation processing on the reconstructed snore signal data, namely segmenting the reconstructed snore signal data into a plurality of sections/frames of snore signal segments.
The method specifically comprises the steps of firstly dividing the reconstructed snore signal data subjected to noise reduction treatment into a plurality of reconstructed snore signal frames, wherein the frame length of each reconstructed snore signal frame is the same; and then calculating the short-time energy and the zero crossing rate of each reconstructed snore signal frame, and setting a short-time energy high threshold, a short-time energy low threshold and a zero crossing rate threshold. The short-time energy high threshold, the short-time energy low threshold and the zero crossing rate threshold can be set in a user-defined mode according to actual conditions. Further, screening out reconstructed snore signal frames with signal energy between an energy high threshold and an energy low threshold to form a first group of snore signal frames, and further screening out snore signal frames with a zero crossing rate higher than a zero crossing rate threshold from the first group of snore signal frames to obtain an initial snore fragment sequence, wherein the initial snore fragment sequence comprises M initial snore fragments. And finally, optimizing the initial snore fragment sequence, namely processing the snore fragments with overlapping phenomena in the initial snore fragment sequence to generate a snore signal segmentation sequence.
S30: and carrying out feature extraction on the snore signal segmentation sequence to generate snore signal features.
Specifically, in this embodiment, mel Frequency Cepstrum Coefficient (MFCC) is adopted to perform feature extraction on the segmented sequence of the snore signal, that is, mel Frequency conversion is performed on each snore signal segment in the segmented sequence of the snore signal, so as to extract two-dimensional Mel Frequency spectrum features:
F mel ＝2595log 10 (1+f/700)
wherein, F mel Is the frequency in Mel and f is the actual frequency in Hz.
It will be understood that this embodiment converts the segmented sequence of snore signals from the original Hz frequency to the Me l frequency. The snore signal is characterized by the Mel frequency. In the sound signal processing process, the frequency domain (Hz frequency) of the sound signal is converted into a perception frequency domain (Mel frequency) to better simulate the auditory process.
Specifically, the step of extracting the MFCC features of the snore signal segmentation sequence comprises the following specific steps: firstly, normalizing the snore signal segmentation sequence, namely dividing the snore signal segmentation sequence by an audio energy peak value in the snore signal segmentation sequence, and limiting the signal between plus or minus 1; then the snore signal segmentation sequence is framed according to a voice recognition mode, in the embodiment, 25ms of each frame and 10ms of frame shift can be adopted, 64-dimensional MFCC features are extracted from each frame, and finally the snore signal segmentation sequence is processed into a 96 × 64 × 6 Mel spectrogram. Wherein the Mel frequency spectrum diagram is an integral trend characteristic of the snore signal characteristic.
S40: and inputting the snore signal characteristics into a preset neural network model, and outputting target signal characteristics.
Specifically, the snore signal characteristics are used as snore data to be identified and input into a preset neural network model, and the neural network model detects and identifies the snore signal characteristics to generate target signal characteristics. The target signal feature is a deep learning feature output by the neural network model based on a Mel frequency spectrogram of the snore signal feature.
The preset neural network model is a neural network model which is trained in advance. Preferably, in this embodiment, the neural network model is a migrated residual neural network model resnet50. The migrated residual neural network model resnet50 is a neural network model that occupies a dominant position in the field of computer sound signal processing, the mel-frequency spectrogram of the snore signal characteristics is input into a preset migrated residual neural network model resnet50, and the migrated residual neural network model resnet50 transforms the mel-frequency spectrogram of the snore signal characteristics into a deep learning characteristic, namely, a target signal characteristic.
As an example, the residual neural network model resnet50 inputs the snore signal characteristics into the characteristic extraction module and then divides the snore signal characteristics into two paths, one path performs maximum pooling on the input characteristic plane, the other path performs average pooling on the input characteristic plane, and then performs pixel-by-pixel (e elementary wise) addition operation on the characteristics of the two paths of pooling layers; then pass through the BatchNorma i zat ion layer, the Leaky ReLU layer and the full-connection layer (the output dimension is 512), finally, only the BatchNorma i zat i on layer is connected behind the full-connection layer, and the activation function layer is not connected, so that the activation function layer is prevented from destroying the output characteristics. Therefore, the loss of the features can be effectively reduced, and the feature extraction capability of the network is enhanced.
S50: and constructing a full connection layer to classify the target signal characteristics and output a snore detection result.
Further, a full connection layer is constructed after the target signal characteristics are output by the neural network model, the preset neural network model is connected with the full connection layer, and the target signal characteristics are input into the full connection layer, so that the target signal characteristics are classified, and the snore detection result is output.
In the embodiment, initial snore signal data is obtained, and noise reduction processing is carried out on the initial snore signal data by adopting a variational modal decomposition method to generate reconstructed snore signal data; adopting a voice endpoint detection method to segment the reconstructed snore signal data to generate a snore signal segmentation sequence; performing feature extraction on the snore signal segmentation sequence to generate snore signal features; inputting the snore signal characteristics into a preset neural network model, and outputting target signal characteristics; constructing a full connection layer to classify the target signal characteristics and outputting snore detection results; in the embodiment, the whole-process snore detection of the snore signals is realized by carrying out variation modal decomposition and noise reduction on the obtained initial snore signal data, carrying out segmentation processing on the reconstructed snore signal data by utilizing a voice endpoint detection method, and extracting and combining a feature extraction method and a transfer learning method; and further, the accuracy of the detection result output when the snore signal is detected is improved.
In an embodiment, as shown in fig. 3, before acquiring initial snore signal data, the snore detecting method specifically includes the following steps:
s11: and acquiring original snore signal data according to a preset sampling frequency and a preset sampling time length.
As an example, the original snore signal data is collected by an audio collecting device according to a preset sampling frequency and a preset sampling time length. The audio acquisition equipment is equipment capable of acquiring sound signals and converting the sound signals into analog audio signals. The audio acquisition device generally includes a plurality of microphones, an audio acquisition controller, and a plurality of analog-to-digital converters. Since the audio acquisition device acquires the original audio signal by using the microphone, the acquired audio signal is an analog audio signal, the analog signal needs to be converted into a digital signal, and the snore signal data processed in the subsequent steps refers to the digital signal acquired here.
In this embodiment, the audio acquisition device is used to acquire raw snore signal data. Wherein, the original snore signal data is the snore signal data without any processing. Specifically, original snore signal data are collected according to a preset sampling frequency and a preset sampling time length. Wherein, the sampling frequency is the number of times of adoption within a certain time range. The sampling time duration is the time duration from the beginning to the end of the acquisition of the original snore signal data. For example: assuming that the collected original snore signal data is 50Hz, the sampling period is 20ms, and the sampling period is primarily set to be 1 period and 200 sampling points, (note: at least 20 points are collected in one period, namely the sampling rate is at least 1 k), the interval of every 2 sampling points is 20ms/200=100us, 100us is collected in one input signal period (20 ms), namely the sampling frequency is 200 times at the moment, and the sampling time length is 20ms.
In the practical application process, because the energy of the snore signal data is mostly distributed within 4kHz and the snore signal data has the characteristic of quasi-periodicity, the snore signal data is collected by the audio collection equipment. For example: the audio acquisition equipment adopts a sampling rate 2 times of the snore energy distribution area when acquiring an original audio signal, namely 8k of sampling frequency, and can adopt a sampling frequency 2-3 times of the snore frequency as the sampling frequency in practice; and snore detection and judgment are carried out every 5 seconds according to the snore period.
S12: and preprocessing the original snore signal data, removing interference and baseline drift, and generating initial snore signal data.
After the raw snore signal data which is not processed is acquired, the raw snore signal data needs to be preprocessed, so that the influence of interference signals or baseline drift existing in the raw snore signal data on the accuracy of a final snore detection result is avoided. In which a baseline drift, which is a curve of a particularly low frequency, is superimposed on the original signal, so that the original signal has a slow, slight tendency to float up and down. The essence of the baseline wander is that a dc component, a low frequency component, is superimposed on the original signal. In the actual signal acquisition process, if the baseline drift/trend term is not eliminated, the trend term is regarded as the true acquired original signal, which affects the accuracy of the signal and the subsequent data processing result. When baseline drift exists, distortion occurs when signal analysis (such as FFT analysis, correlation analysis, power spectral density analysis and the like) is performed subsequently, so that low frequency spikes occur, even main frequency components are submerged, and the precision is seriously affected.
In view of this, the present embodiment preprocesses the collected raw snore signal data to remove interference and baseline drift. Specifically, the interference and baseline drift removal can adopt methods such as least square fitting, wavelet transformation, EMD method, convex optimization, smooth prior method, VMD, FIR filtering, median filtering, low-pass filter and the like. Preferably, because the EMD has a time scale characteristic based on the signal itself, it is not necessary to select a basis function, and it is very suitable for processing nonlinear and non-stationary signals, and has the advantages of high signal-to-noise ratio, etc., and this embodiment uses the EMD method to remove interference and baseline drift.
In the embodiment, original snore signal data are collected according to preset sampling frequency and sampling duration; preprocessing the original snore signal data, removing interference and baseline drift, and generating initial snore signal data; in the embodiment, the acquired original snore signal data is preprocessed in advance, so that the influence of interference signals in the original snore signal data on the accuracy of a subsequent snore detection result can be avoided.
In an embodiment, as shown in fig. 4, in step S10, noise reduction processing is performed on the initial snore signal data by using a variational modal decomposition method to generate reconstructed snore signal data, which specifically includes the following steps:
s101: and carrying out variation modal decomposition on the initial snore signal data to generate K modal components, wherein K is a positive integer.
As an example, performing variational modal decomposition on the preprocessed initial snore signal data x (t) to generate K modal components;
wherein u is k ＝{u 1 ,...,u i ,...,u K Is the decomposed K modal components, u, ordered from high to low frequency i Representing the ith modal component.
S102: an energy entropy of each of the modal components is calculated.
As an example, the energy entropy H (u) of each modal component obtained in step S101 is calculated k )，
H(u k )＝-P k log 2 P k
Wherein, H (u) k ) As modal component u k Energy entropy value of, P k Representing the modal component u k Is in the total energy of the signal x (t), wherein P k Is a known quantity which is preset.
S103: obtaining the ith modal component corresponding to the energy entropy of the local minimum value, filtering the 1 st to ith modal components, and reconstructing the (i + 1) th to Kth modal components to generate reconstructed snore signal data, wherein i is a positive integer smaller than K.
Specifically, the ith modal component u corresponding to the energy entropy of the local minimum value in the energy entropies of the K modal components is obtained i The first 1 to i modal components { u } 1 ,u 2 ,...,u i Filtering, and filtering the residual i + 1-K modal components { u } i+1 ,u i+2 ...,u K Reconstructing the residual i + 1-K modal components { u } i+1 ,u i+2 ...,u K Reconstructing the (i + 1) th to Kth modal components (u) i+1 ,u i+2 ...,u K Recombining to obtain reconstructed snore signal data after noise reduction treatment, wherein i is a positive integer smaller than K.
As an example, if K is 10, that is, 10 modal components are included, then the 6 th modal component u corresponding to the energy entropy of the local minimum value in the energy entropies of the 10 modal components is obtained 6 The first 1 st to 6 th modal components { u } 1 ,u 2 ,...,u 6 Filtering out the rest 7 th to 10 thModal component { u 7 ,..,u 10 H, reconstructing, wherein the remaining 7 th to 10 th modal components { u } 7 ,..,u 10 Reconstructing the 7 th to 10 th modal components { u } 7 ,..,u 10 Recombining to obtain reconstructed snore signal data after noise reduction treatment.
In the embodiment, K modal components are generated by performing variational modal decomposition on the initial snore signal data, wherein K is a positive integer; calculating the energy entropy of each modal component; obtaining an ith modal component corresponding to the energy entropy of the local minimum value, filtering the 1 st to ith modal components, reconstructing the (i + 1) th to Kth modal components to obtain variation modal decomposition, and generating reconstructed snore signal data, wherein i is a positive integer smaller than K; therefore, the effectiveness of noise reduction processing on the initial snore signal data is improved.
In an embodiment, as shown in fig. 5, in step S20, a speech endpoint detection method is used to segment the reconstructed snore signal data to generate a segmented sequence of snore signals, which specifically includes the following steps:
s201: and carrying out preset frame length division on the reconstructed snore signal data to generate a plurality of reconstructed snore signal frames.
The preset frame length is the length of a preset data frame. For example: presetting one frame for every 20ms of snore signals, thereby dividing the reconstructed snore signal data into a plurality of reconstructed snore signal frames. The user can self-define the size of the divided frame length according to the actual situation. It should be noted that, in this embodiment, the frame length of each reconstructed snore signal frame is the same, and each reconstructed snore signal frame is independent.
S202: and calculating the signal energy and the zero crossing rate of each reconstructed snore signal frame, and setting a short-time energy high threshold, a short-time energy low threshold and a zero crossing rate threshold.
Specifically, the signal energy and the zero crossing rate of each reconstructed snore signal frame are calculated, and a short-time energy high threshold value energy _ h, a short-time energy low threshold value energy _ l and a zero crossing rate threshold value zcr are set;
wherein, energy represents the sum of the signal energies of all the reconstructed snore signal frames in the reconstructed snore signal data, and as can be seen from the above formula, the short-time energy high threshold value energy _ h in this embodiment is set to be a quarter of the sum of the signal energies of all the reconstructed snore signal frames in the reconstructed snore signal data.
Wherein energy 1～5 Represents the sum of the signal energies of the first five reconstructed snore signal frames in the reconstructed snore signal data. As can be seen from the above formula, the short-time energy low threshold value energy _ l in this embodiment is set as the sum of the signal energies of the first five reconstructed snore signal frames in the reconstructed snore signal data 1～5 One fifth after addition to the short time energy high threshold energy _ h.
And the zero crossing rate of each reconstructed snore signal frame is the number of zero crossing points of the frame signal in a time domain. In this embodiment, the average mean value of the zero-crossing rates of all the reconstructed snore signal frames in the reconstructed snore signal data is set as the zero-crossing rate threshold, that is, the zero-crossing rates of all the reconstructed snore signal frames in the reconstructed snore signal data are added and divided by the number of the reconstructed snore signal frames to obtain the zero-crossing rate threshold.
S203: and screening out the reconstructed snore signal frames with the signal energy between the energy high threshold and the energy low threshold to form a first group of snore signal frames.
After the short-time energy high threshold value and the short-time energy low threshold value are determined, the signal energy of each reconstructed snore signal frame in the reconstructed snore signal data is compared with the short-time energy high threshold value and the short-time energy low threshold value, so that the reconstructed snore signal frames with the signal energy between the short-time energy high threshold value and the short-time energy low threshold value are screened out, and a first group of snore signal frames are formed. It is understood that the signal energy of all snore signal frames in the first set of snore signal frames is between the short-term energy-high threshold and the short-term energy-low threshold.
S204: screening out snore signal frames with the zero crossing rate higher than the zero crossing rate threshold value from the first group of snore signal frames to obtain an initial snore fragment sequence, wherein the initial snore fragment sequence comprises M initial snore fragments.
Further, after a first group of snore signal frames with signal energy size between the short-time energy high threshold and the short-time energy low threshold are determined, comparing a zero crossing rate of each snore signal frame in the first group of snore signal frames with the zero crossing rate threshold, so as to screen out snore signal frames with zero crossing rates higher than the zero crossing rate threshold from the first group of snore signal frames, and obtain an initial snore fragment sequence. Wherein the initial snore fragment sequence comprises M initial snore fragments. For example: the initial snore fragment sequence S =representsthe first segmented initial snore fragment, and i is smaller than or equal to M. It is understood that the zero crossing rate of all M initial snore fragments is above the zero crossing rate threshold.
S205: and optimizing the initial snore fragment sequence to generate a snore signal segmentation sequence.
Further, in order to avoid that two initial snore fragments at least partially overlapping exist in the initial snore fragment sequence, this embodiment performs optimization processing on the initial snore fragment sequence to generate a snore signal segmentation sequence. As an example, an initial snore fragment in the sequence of initial snore fragments, in which there is at least partial overlap between two adjacent initial snore fragments, may be replaced or deleted.
In this embodiment, the preset frame length division is performed on the reconstructed snore signal data to generate a plurality of reconstructed snore signal frames; calculating the signal energy and the zero crossing rate of each reconstructed snore signal frame, and setting a short-time energy high threshold, a short-time energy low threshold and a zero crossing rate threshold; screening out a reconstructed snore signal frame with signal energy between the short-time energy high threshold and the short-time energy low threshold to form a first group of snore signal frames; screening snore signal frames with the zero crossing rate higher than the zero crossing rate threshold value from the first group of snore signal frames to obtain an initial snore fragment sequence, wherein the initial snore fragment sequence comprises M initial snore fragments; optimizing the initial snore fragment sequence to generate a snore signal segmented sequence; and the reconstructed snore signal data is segmented by adopting an improved voice endpoint detection method, so that the accuracy and the effectiveness of a snore signal segmentation sequence generated after the reconstructed snore signal data is segmented are ensured.
In an embodiment, as shown in fig. 6, in step S205, the optimizing process is performed on the initial snore fragment sequence to generate a snore signal segmentation sequence, which specifically includes the following steps:
s2051: and acquiring any two adjacent snore signal segments from the snore signal segment sequence.
S2052: judging whether the termination time point of the first snore signal segment in the two adjacent snore signal segments is later than the starting time point of the second snore signal segment;
s2053: if the ending time point of the first snore signal segment is later than the starting time point of the second snore signal segment, judging whether the ending time point of the first snore signal segment is later than the ending time point of the second snore signal segment or not;
s2054: if the termination time point of the first snore signal segment is later than the termination time point of the second snore signal segment; and taking the snore signal segment corresponding to the point from the starting time point of the first snore signal segment to the ending time point of the second snore signal segment in the two adjacent snore signal segments as a target snore signal segment, and replacing the two adjacent snore signal segments with the target snore signal segment to generate a snore signal segment sequence.
As an example, the optimizing process of the initial snore fragment sequence specifically includes the following steps: firstly, acquiring any two adjacent snore signal segments from the snore signal segment sequence, then judging whether all overlapping segments exist in the two adjacent snore signal segments, if all overlapping segments exist in the two adjacent snore signal segments, taking the snore signal segment corresponding to the starting time point of the first snore signal segment in the two adjacent snore signal segments to the ending time point of the second snore signal segment as a target snore signal segment, and replacing the two adjacent snore signal segments with the target snore signal segment to generate the snore signal segment sequence.
As an example, whether all the overlapping segments exist in the two adjacent snore signal segments or not is judged, that is, the termination time point of the first snore signal segment in the two adjacent snore signal segments is later than the start time point of the second snore signal segment, and the termination time point of the first snore signal segment is later than the termination time point of the second snore signal segment, that is, the first snore signal segment completely covers the second snore signal segment, at this time, the snore signal segment corresponding to the termination time point from the start time point of the first snore signal segment in the two adjacent snore signal segments to the termination time point of the second snore signal segment is used as a target snore signal segment, and the target snore signal segment is replaced by the two adjacent snore signal segments, so as to generate a sequence of the snore signal segments.
In this embodiment, any two adjacent snore signal segments are obtained from the sequence of snore signal segments; judging whether all overlapping segments exist in any two adjacent snore signal segments; if all the overlapping segments exist in the two adjacent snore signal segments, taking the snore signal segment corresponding to the point from the starting time point of the first snore signal segment to the ending time point of the second snore signal segment in the two adjacent snore signal segments as a target snore signal segment, and replacing the target snore signal segment with the two adjacent snore signal segments to generate a snore signal segment sequence; therefore, the problem that the accuracy and the effectiveness of the generated snore signal segmented sequence are influenced by the fact that all overlapped segments exist in any two adjacent snore signal segments in the snore signal segment sequence can be avoided.
In one embodiment, after determining whether the ending time point of the first snore signal segment is later than the ending time point of the second snore signal segment, the snore detecting method further comprises the steps of:
and if the termination time point of the first snore signal segment is earlier than the termination time point of the second snore signal segment, reserving the first snore signal segment in the two adjacent snore signal segments, deleting the second snore signal segment, and generating a snore signal segment sequence.
Specifically, if the ending time point of the first snore signal segment is earlier than the ending time point of the second snore signal segment, it can be considered that a partial overlapping segment phenomenon exists between the two adjacent snore signal segments.
In this embodiment, if the ending time point of the first snore signal segment is earlier than the ending time point of the second snore signal segment, the first snore signal segment in the two adjacent snore signal segments is retained, and the second snore signal segment is deleted to generate a sequence of snore signal segments, so that the problem that the accuracy and the effectiveness of the generated sequence of snore signal segments are influenced by the existence of partially overlapped segments in any two adjacent snore signal segments in the sequence of snore signal segments can be avoided.
It should be understood that, the sequence numbers of the steps in the foregoing embodiments do not imply an execution sequence, and the execution sequence of each process should be determined by its function and inherent logic, and should not constitute any limitation to the implementation process of the embodiments of the present invention.
In an embodiment, a snore detecting device is provided, and the snore detecting device corresponds to the snore detecting method in the above embodiment one to one. As shown in fig. 7, the snoring detection apparatus includes a first obtaining module 10, a segmentation processing module 20, a feature extraction module 30, a detection module 40, and a classification module 50. The functional modules are explained in detail as follows:
the first acquisition module 10 is configured to acquire initial snore signal data, and perform noise reduction processing on the initial snore signal data by using a variational modal decomposition method to generate reconstructed snore signal data;
a segmentation processing module 20, configured to perform segmentation processing on the reconstructed snore signal data by using a voice endpoint detection method to generate a snore signal segmentation sequence;
the feature extraction module 30 is configured to perform feature extraction on the snore signal segmentation sequence to generate a snore signal feature;
the detection module 40 is used for inputting the snore signal characteristics into a preset neural network model and outputting target signal characteristics;
and the classification module 50 is used for constructing a full connection layer to classify the target signal characteristics and outputting a snore detection result.
Further, as shown in fig. 8, the snore detecting device further includes:
the acquisition module 11 is used for acquiring original snore signal data according to a preset sampling frequency and sampling duration;
and the preprocessing module 12 is used for preprocessing the original snore signal data, removing interference and baseline drift, and generating initial snore signal data.
Further, the first obtaining module 10 includes:
the variational modal decomposition unit is used for carrying out variational modal decomposition on the initial snore signal data to generate K modal components, wherein K is a positive integer;
the first calculation unit is used for calculating the energy entropy of each modal component;
and the reconstruction unit is used for acquiring the ith modal component corresponding to the energy entropy of the local minimum value, filtering the 1 st to ith modal components, reconstructing the (i + 1) th to Kth modal components and generating reconstructed snore signal data, wherein i is a positive integer smaller than K.
Further, the segmentation processing module 20 includes:
the dividing unit is used for dividing the length of a preset frame of the reconstructed snore signal data to generate a plurality of reconstructed snore signal frames;
the second calculation unit is used for calculating the signal energy and the zero crossing rate of each reconstructed snore signal frame and setting a short-time energy high threshold, a short-time energy low threshold and a zero crossing rate threshold;
the first screening unit is used for screening out reconstructed snore signal frames with signal energy between the energy high threshold and the energy low threshold to form a first group of snore signal frames;
a second screening unit, configured to screen, from the first group of snore signal frames, snore signal frames with a zero crossing rate higher than the zero crossing rate threshold value to obtain an initial snore fragment sequence, where the initial snore fragment sequence includes M initial snore fragments;
and the optimization processing unit is used for optimizing the initial snore fragment sequence to generate a snore signal segmentation sequence.
Further, the optimization processing unit includes:
an acquisition subunit: the snore signal segment sequence is used for acquiring any two adjacent snore signal segments from the snore signal segment sequence;
a first judgment subunit: the snore signal segment judging module is used for judging whether the ending time point of the first snore signal segment in the two adjacent snore signal segments is later than the starting time point of the second snore signal segment;
a second judgment subunit: the snore signal segment judging module is used for judging whether the ending time point of the first snore signal segment is later than the ending time point of the second snore signal segment when the ending time point of the first snore signal segment is later than the starting time point of the second snore signal segment;
replacing the subunit: for when the termination time point of said first snore signal segment is later than the termination time point of said second snore signal segment; and taking the snore signal segment corresponding to the point from the starting time point of the first snore signal segment to the ending time point of the second snore signal segment in the two adjacent snore signal segments as a target snore signal segment, and replacing the two adjacent snore signal segments with the target snore signal segment to generate a snore signal segment sequence.
Further, the optimization processing unit further includes:
and the deleting subunit is used for reserving the first snore signal segment in the two adjacent snore signal segments and deleting the second snore signal segment to generate a snore signal segment sequence when the termination time point of the first snore signal segment is earlier than that of the second snore signal segment.
For the specific definition of the snore detecting device, reference may be made to the above definition of the snore detecting method, which is not described herein again. All modules in the snore detecting device can be completely or partially realized by software, hardware and a combination thereof. The modules can be embedded in a hardware form or independent from a processor in the computer device, and can also be stored in a memory in the computer device in a software form, so that the processor can call and execute operations corresponding to the modules.
In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as shown in fig. 9. The computer device includes a processor, a memory, a network interface, and a database connected by a system bus. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device comprises a nonvolatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of an operating system and computer programs in the non-volatile storage medium. The database of the computer device is used to store data used in implementing the above embodiments. The network interface of the computer device is used for communicating with an external terminal through a network connection. The computer program is executed by a processor to implement a snore detecting method.
In an embodiment, a computer device is provided, comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the steps of the snore detecting method in the above embodiments when executing the computer program.
In an embodiment, a computer readable storage medium is provided, on which a computer program is stored, which when executed by a processor, performs the steps of the snore detecting method in the above described embodiments.
It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by hardware instructions of a computer program, which can be stored in a non-volatile computer-readable storage medium, and when executed, can include the processes of the embodiments of the methods described above. Any reference to memory, storage, database, or other medium used in the embodiments provided herein may include non-volatile and/or volatile memory, among others. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically Programmable ROM (EPROM), electrically Erasable Programmable ROM (EEPROM), or flash memory. Volatile memory can include Random Access Memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms such as Static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double Data Rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synch i nk DRAM (SLDRAM), rambus Direct RAM (RDRAM), direct Rambus Dynamic RAM (DRDRAM), and Rambus Dynamic RAM (RDRAM).
It will be apparent to those skilled in the art that, for convenience and brevity of description, only the above-mentioned division of the functional units and modules is illustrated, and in practical applications, the above-mentioned function distribution may be performed by different functional units and modules according to needs, that is, the internal structure of the apparatus is divided into different functional units or modules to perform all or part of the above-mentioned functions.
The above-mentioned embodiments are only used for illustrating the technical solutions of the present invention, and not for limiting the same; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; such modifications and substitutions do not substantially depart from the spirit and scope of the embodiments of the present invention, and are intended to be included within the scope of the present invention.
Claims (10)
1. A snore detecting method is characterized by comprising the following steps:
acquiring initial snore signal data, and performing noise reduction processing on the initial snore signal data by adopting a variational modal decomposition method to generate reconstructed snore signal data;
adopting a voice endpoint detection method to segment the reconstructed snore signal data to generate a snore signal segmentation sequence;
performing feature extraction on the snore signal segmentation sequence to generate snore signal features;
inputting the snore signal characteristics into a preset neural network model, and outputting target signal characteristics;
and constructing a full connection layer to classify the target signal characteristics and output a snore detection result.
2. The snore detecting method of claim 1, wherein prior to obtaining initial snore signal data, the snore detecting method further comprises:
collecting original snore signal data according to a preset sampling frequency and a preset sampling duration;
and preprocessing the original snore signal data, removing interference and baseline drift, and generating the original snore signal data.
3. The snore detecting method of claim 1, wherein denoising the initial snore signal data using a variational modal decomposition method to generate reconstructed snore signal data, comprises:
carrying out variation modal decomposition on the initial snore signal data to generate K modal components, wherein K is a positive integer;
calculating the energy entropy of each modal component;
obtaining the ith modal component corresponding to the energy entropy of the local minimum value, filtering the 1 st to ith modal components, and reconstructing the (i + 1) th to Kth modal components to generate reconstructed snore signal data, wherein i is a positive integer smaller than K.
4. The snore detecting method of claim 1, wherein segmenting the reconstructed snore signal data using a speech endpoint detection method to generate a snore signal segmentation sequence, comprises:
dividing the reconstructed snore signal data into preset frame lengths to generate a plurality of reconstructed snore signal frames;
calculating the signal energy and the zero crossing rate of each reconstructed snore signal frame, and setting a short-time energy high threshold, a short-time energy low threshold and a zero crossing rate threshold;
screening out reconstructed snore signal frames with signal energy between the energy high threshold and the energy low threshold to form a first group of snore signal frames;
screening snore signal frames with the zero crossing rate higher than the zero crossing rate threshold value from the first group of snore signal frames to obtain an initial snore fragment sequence, wherein the initial snore fragment sequence comprises M initial snore fragments;
and optimizing the initial snore fragment sequence to generate a snore signal segmented sequence.
5. The snore detecting method of claim 4, wherein optimizing the initial snore fragment sequence to generate a sequence of snore signal segments comprises:
acquiring any two adjacent snore signal segments from the snore signal segment sequence;
judging whether the termination time point of the first snore signal segment in the two adjacent snore signal segments is later than the starting time point of the second snore signal segment;
if the termination time point of the first snore signal segment is later than the starting time point of the second snore signal segment, judging whether the termination time point of the first snore signal segment is later than the termination time point of the second snore signal segment;
if the termination time point of the first snore signal segment is later than the termination time point of the second snore signal segment; and taking the snore signal segment corresponding to the point from the starting time point of the first snore signal segment to the ending time point of the second snore signal segment in the two adjacent snore signal segments as a target snore signal segment, and replacing the two adjacent snore signal segments with the target snore signal segment to generate a snore signal segment sequence.
6. The snore detecting method of claim 5, wherein after determining whether the point in time at which the first snore signal segment terminates is later than the point in time at which the second snore signal segment terminates, the snore detecting method further comprises:
and if the termination time point of the first snore signal segment is earlier than the termination time point of the second snore signal segment, reserving the first snore signal segment in the two adjacent snore signal segments, deleting the second snore signal segment, and generating a snore signal segment sequence.
7. A snore detecting device, comprising:
the first acquisition module is used for acquiring initial snore signal data, and performing noise reduction processing on the initial snore signal data by adopting a variational modal decomposition method to generate reconstructed snore signal data;
the segmentation processing module is used for carrying out segmentation processing on the reconstructed snore signal data by adopting a voice endpoint detection method to generate a snore signal segmentation sequence;
the characteristic extraction module is used for carrying out characteristic extraction on the snore signal segmented sequence to generate snore signal characteristics;
the detection module is used for inputting the snore signal characteristics into a preset neural network model and outputting target signal characteristics;
and the classification module is used for constructing a full connection layer to classify the target signal characteristics and outputting a snore detection result.
8. A snoring detection apparatus as claimed in claim 7, wherein said snoring detection apparatus further comprises:
the acquisition module is used for acquiring original snore signal data according to preset sampling frequency and sampling duration;
and the preprocessing module is used for preprocessing the original snore signal data, removing interference and baseline drift and generating initial snore signal data.
9. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor when executing the computer program implements a snore detecting method as claimed in any one of claims 1 to 6.
10. A computer-readable storage medium, in which a computer program is stored, which, when being executed by a processor, carries out the snore detecting method according to any one of claims 1 to 6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211457181.1A CN115762551A (en) | 2022-11-21 | 2022-11-21 | Snore detection method and device, computer equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211457181.1A CN115762551A (en) | 2022-11-21 | 2022-11-21 | Snore detection method and device, computer equipment and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115762551A true CN115762551A (en) | 2023-03-07 |
Family
ID=85333747
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211457181.1A Pending CN115762551A (en) | 2022-11-21 | 2022-11-21 | Snore detection method and device, computer equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115762551A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117388835A (en) * | 2023-12-13 | 2024-01-12 | 湖南赛能环测科技有限公司 | Multi-spelling fusion sodar signal enhancement method |
-
2022
- 2022-11-21 CN CN202211457181.1A patent/CN115762551A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117388835A (en) * | 2023-12-13 | 2024-01-12 | 湖南赛能环测科技有限公司 | Multi-spelling fusion sodar signal enhancement method |
| CN117388835B (en) * | 2023-12-13 | 2024-03-08 | 湖南赛能环测科技有限公司 | Multi-spelling fusion sodar signal enhancement method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108831500B (en) | Speech enhancement method, device, computer equipment and storage medium | |
| CN109767783B (en) | Voice enhancement method, device, equipment and storage medium | |
| CN110120224B (en) | Method and device for constructing bird sound recognition model, computer equipment and storage medium | |
| Shah et al. | Time-frequency mask-based speech enhancement using convolutional generative adversarial network | |
| CN108877775A (en) | Voice data processing method, device, computer equipment and storage medium | |
| CN110123367B (en) | Computer device, heart sound recognition method, model training device, and storage medium | |
| CN108922543B (en) | Model base establishing method, voice recognition method, device, equipment and medium | |
| KR20060044629A (en) | Isolating speech signals utilizing neural networks | |
| Pang | Spectrum energy based voice activity detection | |
| WO2019232848A1 (en) | Voice distinguishing method and device, computer device and storage medium | |
| CN117711419B (en) | Intelligent data cleaning method for data center | |
| CN115762551A (en) | Snore detection method and device, computer equipment and storage medium | |
| Labied et al. | An overview of automatic speech recognition preprocessing techniques | |
| CN117037834A (en) | Conference voice data intelligent acquisition method and system | |
| CN111968651A (en) | WT (WT) -based voiceprint recognition method and system | |
| WO2017128910A1 (en) | Method, apparatus and electronic device for determining speech presence probability | |
| Poovarasan et al. | Speech enhancement using sliding window empirical mode decomposition and hurst-based technique | |
| Azami et al. | Automatic signal segmentation using the fractal dimension and weighted moving average filter | |
| CN112216285B (en) | Multi-user session detection method, system, mobile terminal and storage medium | |
| CN111681649B (en) | Speech recognition method, interaction system and achievement management system comprising system | |
| CN116229992A (en) | Voice lie detection method, device, medium and equipment | |
| TWI749547B (en) | Speech enhancement system based on deep learning | |
| Lili et al. | Research on recognition of CHD heart sound using MFCC and LPCC | |
| CN113948088A (en) | Voice recognition method and device based on waveform simulation | |
| Tong et al. | A robust time-frequency decomposition model for suppression of mixed Gaussian-impulse noise in audio signals |
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 | ||
| CB03 | Change of inventor or designer information |
Inventor after: Liu Bing Inventor after: Fang Mengqing Inventor after: Zheng Qian Inventor after: Yan Shuiping Inventor before: Yan Shuiping Inventor before: Liu Bing Inventor before: Zheng Qian |
|
| CB03 | Change of inventor or designer information |