CN116383709B - Signal noise detection method, signal detection model, and readable storage medium - Google Patents
Signal noise detection method, signal detection model, and readable storage medium Download PDFInfo
- Publication number
- CN116383709B CN116383709B CN202310658542.7A CN202310658542A CN116383709B CN 116383709 B CN116383709 B CN 116383709B CN 202310658542 A CN202310658542 A CN 202310658542A CN 116383709 B CN116383709 B CN 116383709B
- Authority
- CN
- China
- Prior art keywords
- signal
- noise
- time window
- signal time
- noise signal
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 132
- 238000003860 storage Methods 0.000 title claims abstract description 12
- 238000000034 method Methods 0.000 claims abstract description 122
- 238000012545 processing Methods 0.000 claims abstract description 84
- 239000011159 matrix material Substances 0.000 claims description 196
- 238000000354 decomposition reaction Methods 0.000 claims description 87
- 230000000737 periodic effect Effects 0.000 claims description 69
- 238000001914 filtration Methods 0.000 claims description 28
- 238000000926 separation method Methods 0.000 claims description 28
- 230000007613 environmental effect Effects 0.000 claims description 26
- 238000005259 measurement Methods 0.000 claims description 13
- 238000005070 sampling Methods 0.000 claims description 9
- 230000000035 biogenic effect Effects 0.000 claims description 6
- 230000002452 interceptive effect Effects 0.000 claims description 5
- 238000004590 computer program Methods 0.000 claims description 3
- 238000011946 reduction process Methods 0.000 abstract description 2
- 230000000052 comparative effect Effects 0.000 description 25
- 230000003183 myoelectrical effect Effects 0.000 description 11
- 210000004556 brain Anatomy 0.000 description 9
- 230000000694 effects Effects 0.000 description 9
- 230000004424 eye movement Effects 0.000 description 8
- 238000004422 calculation algorithm Methods 0.000 description 7
- 238000009826 distribution Methods 0.000 description 7
- 230000005611 electricity Effects 0.000 description 7
- 230000001037 epileptic effect Effects 0.000 description 5
- 238000004088 simulation Methods 0.000 description 5
- 238000010606 normalization Methods 0.000 description 4
- 238000013145 classification model Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 238000007635 classification algorithm Methods 0.000 description 2
- 238000011156 evaluation Methods 0.000 description 2
- 238000012880 independent component analysis Methods 0.000 description 2
- 238000000513 principal component analysis Methods 0.000 description 2
- 230000005856 abnormality Effects 0.000 description 1
- 238000010420 art technique Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000010219 correlation analysis Methods 0.000 description 1
- 230000004807 localization Effects 0.000 description 1
- 230000033001 locomotion Effects 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 238000005096 rolling process Methods 0.000 description 1
- 230000003595 spectral effect Effects 0.000 description 1
- 238000001228 spectrum Methods 0.000 description 1
- 239000013598 vector Substances 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/10—Pre-processing; Data cleansing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
- G06F18/2134—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on separation criteria, e.g. independent component analysis
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T90/00—Enabling technologies or technologies with a potential or indirect contribution to GHG emissions mitigation
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Computational Biology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)
- Measurement Of Velocity Or Position Using Acoustic Or Ultrasonic Waves (AREA)
- Noise Elimination (AREA)
- Signal Processing Not Specific To The Method Of Recording And Reproducing (AREA)
- Testing, Inspecting, Measuring Of Stereoscopic Televisions And Televisions (AREA)
Abstract
The invention discloses a signal noise detection method, a signal detection model and a readable storage medium, wherein the signal noise detection method comprises the following steps: s1, dividing a signal time window; s2, classifying and detecting the signal time windows, wherein the signal time windows are divided into noise signal time windows and normal signal time windows; s3, detecting noise signal types in a noise signal time window; s4, constructing different signal processing models for the noise signal time window according to the noise signal category so as to separate out the corresponding noise signal. The invention adopts the time-space domain combined method to respectively measure, classify, locate and separate the multi-type noise signals in the signals, can accurately measure and separate various noise signals, reduces the loss of effective signals in the noise reduction process and improves the signal-to-noise ratio of the signals.
Description
Technical Field
The invention relates to the technical field of signal detection, in particular to a signal noise detection method, a signal detection model and a readable storage medium.
Background
Because other noise signals are often accompanied in the acquisition process of the original signals, the signal-to-noise ratio of the signals is low, and the signal quality is greatly influenced, the noise reduction is a necessary step of preprocessing before the signals are applied. The noise reduction process generally includes locating and removing noise signals. The noise signal is usually first subjected to positioning analysis by using methods such as ICA (independent component analysis), PCA (principal component analysis technique) and the like, and then removed by adopting a filtering mode. However, these methods have two main disadvantages: firstly, the processing of the decomposed components is too rough, and the whole sub-components containing noise signals are usually directly discarded, so that some useful signals are also removed; and secondly, the type of the noise signal is single, and when the signal contains multiple types of noise signals, the same processing mode is still adopted.
Disclosure of Invention
The invention aims to solve one of the technical problems existing in the prior art. Therefore, the invention provides a signal noise detection method, a signal detection model and a readable storage medium, and the signal to noise ratio of a reference signal can be improved by measuring, classifying, evaluating, positioning and separating noise signals in an original signal.
The technical scheme adopted for solving the technical problems is as follows: a method of detecting signal noise, comprising: dividing a signal time window; classifying and detecting the signal time window, wherein the signal time window is divided into a noise signal time window and a normal signal time window; measuring a noise signal class in the noise signal time window; and constructing different signal processing models for the noise signal time window according to the noise signal category so as to separate out the corresponding noise signal.
The invention has the beneficial effects that the invention can realize the measurement and separation of multiple types of noise signals, different noise signals have corresponding signal processing models, can improve the accuracy of noise signal removal, and can reduce the loss of effective signals.
Further, the signal processing model includes: acquiring a time decomposition matrix V and/or a space decomposition matrix U based on the noise signal time window, wherein the time decomposition matrix V is a K multiplied by K matrix, and K is the number of time domain sampling points; the space decomposition matrix U is an M×M matrix, and M is the channel number; extracting the first M columns from the time decomposition matrix V to form a source component matrix A, wherein M is the channel number; positioning the position of the noise signal in the source component matrix A; judging whether the position of a noise signal in the source component matrix A is a null value, and when the position of the noise signal is a non-null value, sequentially carrying out noise signal separation and source reconstruction so as to continue iteration; and when the position of the noise signal is null, ending iteration.
Further, the source reconstruction includes: outputting a matrix V_hat after filtering the noise signals; and performing source decomposition inverse operation on the space decomposition matrix U and the matrix V_hat to obtain a signal S_hat.
Further, measuring the noise signal class in the noise signal time window includes: detecting a non-biological source noise signal and a biological source noise signal in the noise signal; setting a priority order; and sequentially operating different signal processing models according to the priority levels so as to separate noise signals of different levels step by step.
Further, the priority order between the non-biological source noise signals and the biological source noise signals is set according to the energy of the noise signals; the priority order between the non-biological source noise signals and the biological source noise signals is as follows: non-biogenic noise signal > biogenic noise signal.
Further, the non-biological source noise signal comprises: power frequency noise, environmental electromagnetic interference, step noise.
Further, the priority order of the non-biological source noise signals is as follows: the power frequency noise is greater than the environmental electromagnetic interference and the step noise.
Further, the measuring of the power frequency noise and the environmental electromagnetic interference includes: positioning the position of power frequency noise and environmental electromagnetic noise signals through the space characteristics and the time-frequency characteristics; the separation of the power frequency noise and the environmental electromagnetic interference comprises the following steps: and filtering the source components containing the power frequency noise and the environmental electromagnetic interference at a specific frequency to separate the power frequency noise and the environmental electromagnetic noise signals.
Further, the step noise measurement includes: performing first-order difference processing on each column of the source component matrix A, and determining the position of a step noise signal according to the absolute value of the difference value; the step noise separation includes: and performing wavelet threshold filtering processing on the components corresponding to the step noise signals to separate out the step noise signals.
Further, the biological source noise signal includes: at least one of a high frequency noise signal, a low frequency noise signal, and a periodic noise signal.
Further, the priority order of the biological source noise signals is as follows: high frequency noise signal > low frequency noise signal > periodic noise signal.
Further, the measuring of the high frequency noise signal includes: calculating the energy characteristics of each column of signals in the source component matrix A; calculation ofHigh-low frequency energy ratio r of each column signal hl The method comprises the steps of carrying out a first treatment on the surface of the If the high-low frequency energy ratio r hl > set energy ratio threshold r 0 Marking the corresponding column signal as high-frequency noise signal subcomponent, and setting the set of all high-frequency noise signal subcomponent positions as the position L of the high-frequency noise signal h 。
Further, the separation of the high frequency noise signal includes: to position L h The high-frequency interference subcomponent is subjected to filtering processing or CCA filtering processing to separate out a high-frequency noise signal.
Further, the measuring of the low frequency noise signal includes: performing standardization processing on the source component matrix A to obtain a source component matrix B; carrying out peak statistics on each column of signals in the source component matrix B to obtain the peak value number of each column of signals; calculating the peak frequency p of each column of signals in unit time m1 The method comprises the steps of carrying out a first treatment on the surface of the If the peak frequency p m1 > set frequency threshold p 0 The corresponding column signal is marked as a low-frequency noise signal sub-component, and the time position L of the low-frequency noise signal sub-component is recorded v The method comprises the steps of carrying out a first treatment on the surface of the Acquiring a space decomposition matrix U based on the noise signal time window, wherein the space decomposition matrix U is an M multiplied by M matrix, and M is the channel number; each column of signals of the space decomposition matrix U is mapped and then matched with a space position template to obtain a matching coefficient; if the matching coefficient is greater than the set matching threshold, outputting the spatial position L of the low-frequency noise signal subcomponent u The method comprises the steps of carrying out a first treatment on the surface of the According to formula L l =merge(L v ,L u ) Determining the position L of a low-frequency noise signal l 。
Further, the separation of the low frequency noise signal includes: to position L l The low-frequency noise signal subcomponents are subjected to wavelet threshold filtering processing to separate out the low-frequency noise signal.
Further, the measuring of the periodic noise signal includes: performing standardization processing on the source component matrix A to obtain a source component matrix C; carrying out peak statistics on each column of signals in the source component matrix C to obtain the peak value number of each column of signals; calculating the peak frequency p of each column of signals in unit time m2 The method comprises the steps of carrying out a first treatment on the surface of the If the peak frequency p m2 > set frequency threshold p 0 The corresponding column signal is marked as a periodic noise signal sub-component, and the position of the periodic noise signal sub-component is set as the position L of the periodic noise signal z 。
Further, the separation of the periodic noise signal includes: the position L is matched by a template matching method z The periodic interfering subcomponents are processed to separate out a periodic noise signal.
Further, detecting the signal time window by using a time window detection method; the time window detection method comprises the following steps: at least one of generalized characteristic decomposition method, multi-characteristic classification method and multi-characteristic joint detection method.
Further, the generalized characteristic decomposition method includes: respectively calculating covariance matrixes of the signal time windows and the background signals; calculating the maximum generalized eigenvalue between the covariance matrix of the signal time window and the covariance matrix of the background signal; if the maximum generalized characteristic value is larger than the set threshold value, marking the signal time window as a noise signal time window; otherwise, the normal signal time window is marked.
Further, the multi-feature joint detection method comprises: calculating a plurality of characteristic values of the signal time window; and judging the threshold value of the plurality of characteristic values one by one, and marking the signal time window as a noise signal time window when the plurality of characteristic values of the signal time window meet the threshold value condition.
Further, the multi-feature classification method includes: and calculating a plurality of characteristic values of the signal time window, inputting the characteristic values into a classifier for classification, and outputting a noise signal time window and a normal signal time window.
Further, measuring the noise signal class in the noise signal time window further includes: dividing the noise signal time window into the following steps according to the space-time domain sparseness degree of the noise signal: sparse noise signal time window and dense noise signal time window.
Further, the dividing signal time window comprises at least two stages of division, namely a first-stage signal time window is formed by a plurality of second-stage signal time windows; the classifying detection of the signal time window further comprises: and judging a primary signal time window based on the detection result of the secondary signal time window in a combined mode so as to divide the primary signal time window into a sparse noise signal time window, a dense noise signal time window and a normal signal time window.
Further, the step of jointly judging the primary signal time window based on the detection result of the secondary signal time window comprises the following steps: detecting a secondary signal time window, which is divided into a secondary interference time window and a secondary normal time window; calculating the duty ratio p of a second-level interference time window in a first-level signal time window, namely marking the first-level signal time window as a normal signal time window when the duty ratio p=0; marking the primary signal time window as a sparse noise signal time window when the duty cycle 0<p < duty cycle threshold; and when the duty ratio p is more than or equal to the duty ratio threshold value, marking the primary signal time window as a dense noise signal time window.
Further, when the primary signal time window is a sparse noise signal time window, constructing different signal processing models only for a secondary interference time window in the primary signal time window; when the primary signal time window is a dense noise signal time window, different signal processing models are constructed for the primary signal time window.
The invention also provides a signal detection model of the signal noise detection method, which comprises the following steps: the signal time window dividing model is used for dividing the signal time window; the signal time window detection model is used for carrying out classification detection on the signal time windows to obtain noise signal time windows and normal signal time windows; the classification model is used for classifying noise signal categories in the noise signal time window; and the signal processing model separates noise signals of different categories according to the noise signal category.
The present invention also provides a computer readable storage medium having a computer program stored thereon, which when executed by a processor implements the signal noise detection method.
The invention also provides a signal time window detection model based on the generalized characteristic decomposition method, which comprises the following steps: covariance matrix of signal time window and covariance matrix of background signal, and maximum generalized eigenvalue between them and its set threshold; the signal time window detection model runs the signal noise detection method, so that the signal time window is divided into a noise signal time window and a normal signal time window according to a generalized characteristic decomposition method.
The invention also provides a signal time window detection model based on the multi-feature classification method, which comprises the following steps: a plurality of characteristic values of the signal time window and corresponding threshold conditions; the signal time window detection model runs the signal noise detection method, so that the signal time window is divided into a noise signal time window and a normal signal time window according to the multi-feature joint detection method.
The invention also provides a signal time window detection model based on the multi-feature joint detection method, which comprises the following steps: a plurality of characteristic values of the classifier and the signal time window; the signal time window detection model runs the signal noise detection method, so that the signal time window is divided into a noise signal time window and a normal signal time window according to the multi-feature classification method.
The invention also provides a high-frequency noise signal detection model, which comprises the following steps: a signal time window dividing model for dividing the signal time window; the signal time window detection model divides the signal time window into a noise signal time window and a normal signal time window; and the signal processing model runs the signal noise detection method to separate the high-frequency noise signals in the noise signal time window.
Further, the signal processing model includes: acquiring a time decomposition matrix V based on the noise signal time window, wherein the time decomposition matrix V is a K multiplied by K matrix, and K is the number of time domain sampling points; extracting the first M columns from the time decomposition matrix V to form a source component matrix A, wherein M is the channel number; positioning the position of the high-frequency noise signal in the source component matrix A; judging whether the position of the high-frequency noise signal in the source component matrix A is a null value or not; when the position of the high-frequency noise signal is a non-null value, sequentially carrying out separation of the high-frequency noise signal and source reconstruction so as to continue iteration; and when the position of the high-frequency noise signal is null, ending iteration.
Further, the positioning the position of the high-frequency noise signal in the source component matrix a includes: calculating the energy characteristics of each column of signals in the source component matrix A; calculating the high-low frequency energy ratio of each column of signalsr hl The method comprises the steps of carrying out a first treatment on the surface of the If the high-low frequency energy ratio r hl > set energy ratio threshold r 0 Marking the corresponding column signal as high-frequency noise signal subcomponent, and setting the set of all high-frequency noise signal subcomponent positions as the position L of the high-frequency noise signal h 。
Further, the separating the high frequency noise signal includes: to position L h The high-frequency interference subcomponent is subjected to low-pass filtering processing or CCA filtering processing to separate out a high-frequency noise signal.
The invention also provides a low-frequency noise signal detection model, which comprises the following steps: a signal time window dividing model for dividing the signal time window; the signal time window detection model divides the signal time window into a noise signal time window and a normal signal time window; and the signal processing model runs the signal noise detection method to separate the low-frequency noise signals in the noise signal time window.
Further, the signal processing model includes: acquiring a space decomposition matrix U and a time decomposition matrix V based on the noise signal time window, wherein the space decomposition matrix U is an M×M matrix, M is the channel number, the time decomposition matrix V is a K×K matrix, and K is the time domain sampling point number; extracting the first M columns from the time decomposition matrix V to form a source component matrix A, wherein M is the channel number; positioning the position of a low-frequency noise signal in the source component matrix A; judging whether the position of the low-frequency noise signal in the source component matrix A is a null value or not; when the position of the low-frequency noise signal is a non-null value, sequentially carrying out separation low-frequency noise signal and source reconstruction so as to continue iteration; and when the position of the low-frequency noise signal is null, ending iteration.
Further, the positioning the position of the low-frequency noise signal in the source component matrix a includes: performing standardization processing on the source component matrix A to obtain a source component matrix B; carrying out peak statistics on each column of signals in the source component matrix B to obtain the peak value number of each column of signals; calculating the peak frequency p of each column of signals in unit time m1 The method comprises the steps of carrying out a first treatment on the surface of the If the peak frequency p m1 > set frequency threshold p 0 The corresponding column signal is marked as a low-frequency noise signal sub-component, and the low-frequency noise signal sub-component is recordedTime position L of (2) v The method comprises the steps of carrying out a first treatment on the surface of the Matching the mapped residual space position templates of each column of signals of the space decomposition matrix U to obtain a matching coefficient; if the matching coefficient is larger than the set matching threshold, outputting the space position Lu of the low-frequency noise signal subcomponent; according to formula L l =merge (Lv, lu) to determine the position L of the low-frequency noise signal l 。
Further, the separating the low frequency noise signal includes: to position L l The low-frequency interference subcomponent is subjected to wavelet threshold filtering processing to separate out a low-frequency noise signal.
The invention also provides a periodic noise signal detection model, comprising: a signal time window dividing model for dividing the signal time window; the signal time window detection model is used for dividing the signal time window into a noise signal time window and a normal signal time window; and the signal processing model runs the signal noise detection method to separate periodic noise signals in the noise signal time window.
Further, the method comprises the steps of: the signal processing model includes: acquiring a time decomposition matrix V based on the noise signal time window, wherein the time decomposition matrix V is a K multiplied by K matrix, and K is the number of time domain sampling points; extracting the first M columns from the time decomposition matrix V to form a source component matrix A, wherein M is the channel number; positioning the position of the periodic noise signal in the source component matrix A; judging whether the position of the periodic noise signal in the source component matrix A is a null value or not; when the position of the periodic noise signal is a non-null value, sequentially carrying out separation periodic noise signal and source reconstruction so as to continue iteration; and when the position of the periodic noise signal is null, ending iteration.
Further, the locating the position of the periodic noise signal in the source component matrix a includes: performing standardization processing on the source component matrix A to obtain a source component matrix C; carrying out peak statistics on each column of signals in the source component matrix C to obtain the peak value number of each column of signals; calculating the peak frequency p of each column of signals in unit time m2 The method comprises the steps of carrying out a first treatment on the surface of the If the peak frequency p m2 > set frequency threshold p 0 The corresponding column signal is marked as a periodic noise signal sub-component, which is set The position of the sub-component of the periodic noise signal is the position L of the periodic noise signal z 。
Further, the separation of the periodic noise signal includes: the position L is matched by a template matching method z The periodic interfering subcomponents are processed to separate out a periodic noise signal.
The invention also provides a multi-stage detection model of the signal time window, which comprises the following steps: the signal time window dividing model divides the signal time windows in at least two stages, namely a plurality of two-stage signal time windows form a first-stage signal time window; the signal time window detection model is used for running the signal noise detection method, and the secondary signal time window is divided into a secondary interference time window and a secondary normal time window by the time window detection method; and the classification model is used for jointly judging the primary signal time window based on the detection result of the secondary signal time window so as to classify the primary signal time window into a sparse noise signal time window, a dense noise signal time window and a normal signal time window.
Further, the time window detection method is at least one of a generalized characteristic decomposition algorithm, a multi-characteristic classification algorithm and a multi-characteristic joint detection algorithm.
Further, the generalized characteristic decomposition method includes: calculating the characteristic matrix of the signal time window and the background signal respectively; calculating the maximum generalized eigenvalue between the eigenvalue of the signal time window and the eigenvalue of the background signal; if the maximum generalized characteristic value is larger than the set threshold value, marking the signal time window as a noise signal time window; otherwise, the normal signal time window is marked.
Further, the multi-feature joint detection method comprises: calculating a plurality of characteristic values of the signal time window; and judging the threshold value of the plurality of characteristic values one by one, and marking the signal time window as a noise signal time window when the plurality of characteristic values of the signal time window meet the threshold value condition.
Further, the multi-feature classification method includes: and calculating a plurality of characteristic values of the signal time window, inputting the characteristic values into a classifier for classification, and outputting a noise signal time window and a normal signal time window.
Further, the step of jointly judging the primary signal time window based on the detection result of the secondary signal time window includes: calculating the duty ratio p of a second-level interference time window in the first-level signal time window; when the duty ratio p=0, marking the primary signal time window as a normal signal time window; marking the primary signal time window as a sparse noise signal time window when the duty cycle 0<p < duty cycle threshold; and when the duty ratio p is more than or equal to the duty ratio threshold value, marking the primary signal time window as a dense noise signal time window.
Drawings
The invention will be further described with reference to the drawings and examples.
Fig. 1 is a flowchart of a signal noise detection method of the present invention.
Fig. 2 is a flow chart of a signal noise detection method according to the present invention according to priority processing.
Fig. 3 is a flow chart of the low frequency noise signal localization of the present invention.
Fig. 4 is a graph showing the correlation coefficients of example 1 and comparative examples 1 and 2.
Fig. 5 is a comparative graph of the relative root mean square error of example 1, comparative examples 1, 2.
Fig. 6 is a graph showing the effect of the interference cancellation treatment in example 2 and comparative examples 3 and 4.
Fig. 7 is a schematic diagram of the classification detection of the signal window according to the present invention.
Fig. 8 is a flow chart of the generalized feature decomposition method of the present invention.
FIG. 9 is a flow chart of the multi-feature joint detection of the present invention.
FIG. 10 is a flow chart of multi-feature classification detection of the present invention.
Fig. 11 is a graph of the result of denoising a signal using a generalized eigendecomposition method of the present invention.
Fig. 12 is a graph of the result of a prior art denoising signal.
FIG. 13 is a graph comparing results of identifying noise signals for feature joint detection, single feature detection.
Detailed Description
The invention will now be described in further detail with reference to the accompanying drawings. The drawings are simplified schematic representations which merely illustrate the basic structure of the invention and therefore show only the structures which are relevant to the invention.
In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", "length", "width", "thickness", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", "clockwise", "counterclockwise", "axial", "radial", "circumferential", etc. indicate orientations or positional relationships based on the orientations or positional relationships shown in the drawings are merely for convenience in describing the present invention and simplifying the description, and do not indicate or imply that the device or element being referred to must have a specific orientation, be configured and operated in a specific orientation, and therefore should not be construed as limiting the present invention. Furthermore, features defining "first", "second" may include one or more such features, either explicitly or implicitly. In the description of the present invention, unless otherwise indicated, the meaning of "a plurality" is two or more.
In the description of the present invention, it should be noted that, unless explicitly specified and limited otherwise, the terms "mounted," "connected," and "connected" are to be construed broadly, and may be either fixedly connected, detachably connected, or integrally connected, for example; can be mechanically or electrically connected; can be directly connected or indirectly connected through an intermediate medium, and can be communication between two elements. The specific meaning of the above terms in the present invention will be understood in specific cases by those of ordinary skill in the art.
As shown in fig. 1 to 2, the signal noise detection method of the present invention includes the steps of: s1, dividing a signal time window; s2, classifying and detecting the signal time windows, wherein the signal time windows are divided into noise signal time windows and normal signal time windows; s3, dividing noise signal categories in a noise signal time window; s4, constructing different signal processing models for the noise signal time window according to the noise signal category so as to separate out the corresponding noise signal. It should be noted that, divide the signal into a plurality of signal time windows, and then classify each signal time window again, can improve detection efficiency, after detecting noise signal time window, can obtain noise signal category that noise signal time window contained, different categories construct different signal processing models and separate, more pertinence.
For example, the signal processing model includes: acquiring a time decomposition matrix V and a space decomposition matrix U based on the noise signal time window, wherein the time decomposition matrix V is a K multiplied by K matrix, and K is the number of time domain sampling points; the space decomposition matrix U is an M×M matrix, and M is the channel number; the first M columns are extracted from the time-resolved matrix V to form the source component matrix a, where M is the number of channels. Positioning the position of the noise signal in the source component matrix A; judging whether the position of the noise signal in the source component matrix A is null, and when the position of the noise signal is not null, sequentially carrying out noise signal separation and source reconstruction so as to continue iteration; when the position of the noise signal is null, the iteration is terminated. Wherein the source reconstruction comprises: outputting a matrix V_hat after filtering the noise signals; and performing source decomposition inverse operation on the space decomposition matrix U and the matrix V_hat to obtain a signal S_hat. Of course, the signal processing model in the present case may not iterate, i.e. separate noise signals once.
The time-resolved matrix V is a decomposition of the signal in the time dimension, and each column in the time-resolved matrix V can reflect time-domain distribution information of each sub-component. The spatial decomposition matrix U is a decomposition of the signal in the spatial distribution dimension of the electrode, and each column can reflect spatial distribution information of each sub-component. The time decomposition matrix V and the space decomposition matrix U of the noise signal window may be obtained by a Singular Value Decomposition (SVD) method or a feature decomposition (ED) method, which is not limited herein. Wherein the Singular Value Decomposition (SVD) comprises: decomposing the multichannel signal matrix S into: ，S∈ R M×K Is a signal matrix with M channels and K time samples,/for each channel>∈R M×M Is a left singular matrix, ">∈R M×K Is a singular value matrix, ">∈R K×K Is a right singular matrix. The left singular U matrix is the decomposition of the signal matrix S in the space dimension, each column can reflect the spatial distribution of each sub-component, and the right singular V matrix is the decomposition of the signal matrix S in the time dimension, and each column of V can reflect the time domain distribution of each sub-component. The feature decomposition method (ED) includes: for signal matrix S epsilon R M×K SVD is carried out to obtain a left singular matrix U and a right singular matrix V, and the SVD can be obtained through the following ED equivalence: calculating the spatial covariance matrix rs=s×s of S T ED for Rs: rs= =>U represents a feature matrix of Rs, namely a left singular matrix of S; calculating a time domain covariance matrix rt=s of S T * S/M, ED is carried out on Rt: rt= =>V represents the feature matrix of Rt, namely the right singular matrix of S. The invention constructs a source signal (namely a source component matrix A) before locating the position of a noise signal, wherein A= [ v ] 1 ，v 2 ，...，v M ]，v 1 ～v M Representing the first M columns of the time decomposition matrix V.
The noise signal category in the noise signal time window is measured, including: the non-biological source noise signal and the biological source noise signal in the noise signal are measured. In order to improve the accuracy of noise signal removal, the invention adopts different measuring and separating modes for different types of noise signals. Because the biological source noise signals and the non-biological source noise signals are different in sources, the influence degree on the signals is different, the presence of multiple types of noise signals can influence the signal detection, especially when noise signals with strong energy exist, the noise signals are removed preferentially or simultaneously, and the accurate separation between the signals and the interference sources is facilitated. The non-biological source noise signal and biological source noise signal are determined based on energy (signal to noise ratio), for example and without limitation, and noise signals of definite source are separated first, i.e. noise signals with strong energy are preferentially processed. When the noise signal energy of the non-biological source noise signal far exceeds the biological source noise signal, the non-biological source noise signal is processed preferentially, and then the biological source noise signal is processed, the priority order can be set as follows: non-biogenic noise signal > biogenic noise signal. When the noise signal energy of the non-biological source noise signal is far lower than that of the biological source noise signal, the biological source noise signal is preferentially processed, and then the non-biological source noise signal is processed, and the priority order can be set as follows: non-biological source noise signal < biological source noise signal. And then sequentially operating different signal processing models according to the priority levels so as to separate noise signals with different levels step by step. Any type of noise signal may be removed simultaneously or randomly when the noise signal energy of the non-biological source noise signal is similar to the biological source noise signal.
As an alternative embodiment of the noise signal of non-biological origin.
In this case, non-biological source noise signals include, for example and without limitation: power frequency noise, environmental electromagnetic interference, step noise. Alternatively, any type of noise signal may be removed simultaneously or randomly or in a priority order. Preferably, based on the spatial decomposition matrix, the power frequency noise is generally unavoidable and distributed in all channels, so that the power frequency noise is generally removed first, the environmental electromagnetic interference is also distributed in all channels, so that the noise is secondarily removed, and the step noise is generally generated in a few channels, and the amplitude is greatly changed and finally processed. The priority order of the non-biological source noise signals can be set according to the universality of the noise signals, namely: the power frequency noise is greater than the environmental electromagnetic interference and the step noise.
(1) The treatment of power frequency noise and environmental electromagnetic interference is as follows:
for example, the measurement of power frequency noise and environmental electromagnetic interference includes: and positioning the position of the power frequency noise and the environmental electromagnetic noise signal through the space characteristics and the time-frequency characteristics. The separation of power frequency noise and environmental electromagnetic interference comprises: and filtering the source components containing the power frequency noise and the environmental electromagnetic interference at a specific frequency to separate the power frequency noise and the environmental electromagnetic noise signals. The power frequency noise and the environment electromagnetic interference can influence the signals of all channels, so that the variance of the space decomposition vectors corresponding to the power frequency noise and the environment electromagnetic interference is almost zero. In addition, the frequency spectrum of the time decomposition matrix corresponding to the power frequency noise and the environmental electromagnetic interference has peaks at specific frequencies, so that the positions of the power frequency noise and the environmental electromagnetic interference can be positioned through the spatial characteristics and the time-frequency characteristics. Meanwhile, the power frequency noise and the environmental electromagnetic noise signals can be separated through filtering processing of specific frequencies. For example, the specific frequency of the power frequency noise is 50Hz or 60Hz, and the specific frequency of the environmental electromagnetic interference is generally set according to the electromagnetic frequency in the acquisition environment.
(2) The processing for step noise is as follows:
for example, the step noise measurement includes: and performing first-order difference processing on each column of the source component matrix A, and determining the position of the step noise signal according to the absolute value of the difference value. The step noise separation includes: and performing wavelet threshold filtering processing on the components corresponding to the step noise signals to separate out the step noise signals. Step noise generally causes instantaneous amplitude changes to signals of some channels, so that a first-order difference process is performed on each column of the source component matrix a of the signals, and a position where the absolute value of the difference value far exceeds that of other columns is found, where the step noise is located.
As an alternative embodiment of the bio-sourced noise signal.
In this case, the biological source noise signal includes: at least one of a high frequency noise signal, a low frequency noise signal, and a periodic noise signal. Biological source signals generally include: physiological signals such as myoelectricity, electrooculogram, moving electricity, electrocardio and electroencephalogram are selected as reference signals because myoelectricity belongs to high-frequency signals, electrooculogram and moving electricity belongs to low-frequency signals and electrocardio belongs to periodic signals, and the types of biological source noise signals are different. If myoelectricity is used as a reference signal, the biological source noise signal comprises: at least one of a high-frequency noise signal, a low-frequency noise signal, and a periodic noise signal other than myoelectricity; and so on. For non-brain electricity as a reference signal, a corresponding signal is directly obtained according to a single biological source noise signal measuring or separating method, biological source noise signals except the reference signal can be removed in a biological source noise signal positioning or separating mode, and the reference signal is indirectly obtained, so that the denoising effect is achieved.
If the brain electricity is used as a reference signal, a high-frequency noise signal such as but not limited to an electromyographic signal, a low-frequency noise signal such as but not limited to an electrooculographic signal, a motion electric signal and a periodic noise signal such as but not limited to an electrocardiosignal; running the corresponding signal processing models simultaneously or randomly or in order of priority removes any type of noise signals. For example, since the high-frequency noise signal has wide frequency distribution and wide spatial distribution, other noise signals are easy to recognize abnormality, and the high-frequency noise signal is preferentially processed; the amplitude of the low-frequency noise signal is larger, the space characteristics are obvious, the positioning is relatively easy, and the priority is arranged behind the high-frequency noise signal; and finally, removing the periodic noise signal. The biological source noise signals may therefore be prioritized as: high frequency noise signal > low frequency noise signal > periodic noise signal. The specific processing of the biological source noise signal of the brain electricity comprises the following steps:
(1) The processing for the high frequency noise signal is as follows:
the measurement of the high frequency noise signal includes: calculating the energy characteristics of each column of signals in the source component matrix A, and calculating the high-low frequency energy ratio r of each column of signals hl If the high-low frequency energy ratio r hl > set energy ratio threshold r 0 Marking the corresponding column signal as high-frequency noise signal subcomponent, and setting the set of all high-frequency noise signal subcomponent positions as the position L of the high-frequency noise signal h . The energy characteristic is, for example, the power spectral density psd (m) =pwelch (v) m ) High-low frequency energy ratio of each column signal:
where m=1, 2,..m, h1 and h2 represent the lower and upper bounds, respectively, of the high frequency band, and l1 and l2 represent the lower and upper bounds, respectively, of the low frequency band. If the high-low frequency energy ratio r of a certain column hl ＞r 0 Marking the column signal as a high-frequency noise signal sub-component, and the position L of the high-frequency noise signal h I.e. the set of all high frequency noise signal subcomponent positions. Position L for obtaining high-frequency noise signal h Thereafter, the separation of the high frequency noise signal includes: to position L h The above high-frequency interference subcomponents are subjected to a filtering process or a CCA (typical correlation analysis) filtering process to separate out a high-frequency noise signal.
(2) The processing for the low frequency noise signal is as follows:
the measurement of the low frequency noise signal includes: carrying out standardization treatment on the source component matrix A to obtain a source component matrix B; carrying out peak statistics on each column of signals in the source component matrix B to obtain the peak value number of each column of signals; calculating the peak frequency p of each column of signals in unit time m1 The method comprises the steps of carrying out a first treatment on the surface of the If the peak frequency p m1 > set frequency threshold p 0 The corresponding column signal is marked as a low-frequency noise signal sub-component, and the time position L of the low-frequency noise signal sub-component is recorded v The method comprises the steps of carrying out a first treatment on the surface of the Acquiring a space decomposition matrix U based on a noise signal time window, wherein the space decomposition matrix U is an M multiplied by M matrix, and M is the channel number; each column of signals of the space decomposition matrix U is mapped and then matched with a space position template to obtain a matching coefficient; if the matching coefficient is greater than the set matching threshold, outputting the spatial position L of the low-frequency noise signal subcomponent u The method comprises the steps of carrying out a first treatment on the surface of the According to formula L l =merge(L v ,L u ) Determining the position L of a low-frequency noise signal l 。
The normalization of the source component matrix a is based on the mean μ of the samples m =mean（v m ) Sum of variances sigma m =std（v m ) Proceeding, m=1, 2..m, M represents the number of channels, the formula for the normalization process is:
k=1, 2,..k, K represents the number of time samples, a represents the source component matrix before normalization and B represents the source component matrix after normalization. The method has high generalization after standardization, and the fixed threshold is suitable for different individuals. And (3) carrying out peak detection on each column of signals of the source component matrix B, wherein the peak detection is carried out based on three characteristics of peak height, half-peak width and peak-peak distance. After the number of peaks of each column of signals is obtained, the peak frequency p appearing in unit time is calculated m1 If the peak frequency p of a certain column of signals m1 ＞p 0 The column signal is considered to contain a low-frequency noise signal, the column signal is marked as a low-frequency noise signal sub-component, and the time position L of the low-frequency noise signal sub-component is recorded v . The measurement of the low frequency noise signal also requires joint spatial location to be finally confirmed.
The separation of the low frequency noise signal includes: to position L l The low-frequency noise signal subcomponents are subjected to wavelet threshold filtering processing to separate out the low-frequency noise signal. The wavelet threshold filtering process comprises the following steps: wavelet decomposition is carried out on the subcomponent of the low-frequency noise signal to obtain a wavelet signal q, a threshold value is set for each layer of wavelet signal, the part exceeding the threshold value in each layer of wavelet signal is reserved (marked as p, the part is the noise signal), then the reserved part p is subtracted from the wavelet signal q, the clean wavelet signal can be obtained, and finally the clean wavelet signal is reconstructed to obtain the subcomponent after the low-frequency noise signal is removed.
Taking the electro-oculogram noise signal as an example, as shown in FIG. 3, the time position L of the sub-component of the electro-oculogram noise signal is obtained v Then, it is also necessary to obtain the spatial position L of the electro-oculogram noise signal u . Firstly, a spatial position template related to the electrooculogram is required to be obtained, wherein the spatial position template comprises a blink template, a vertical eye movement template and a horizontal eye movement template, and the blink template, the vertical eye movement template and the horizontal eye movement template can respectively detect the spatial blink position, the spatial vertical eye movement position and the spatial horizontal eye movement position of signals to obtain . Each column U of the spatial decomposition matrix U m Mapping the channel position into a brain terrain space, and then respectively matching with three spatial position templates to obtain three matching coefficients BL (M), VE (M) and HE (M), wherein the number of each matching coefficient is M (the same as the number of channels). Setting a matching threshold of each matching coefficient, and recording u corresponding to the matching coefficient larger than the matching threshold respectively m Three spatial positions are finally output: spatial blink position L u1 Spatial vertical eye movement position L u2 Spatial horizontal eye movement position L u3 . Finally according to formula L l =merge(L v ,L u1 ,L u2 ,L u3 )=L v ∩(L u1 L u2 />L u3 ) The location of the final electro-oculogram noise signal is determined. The mode of determining the position of the low-frequency noise signal by combining the time domain and the space domain can restrict the number of the positions of the low-frequency noise signal and prevent the loss of signals caused by excessive subsequent processing components.
(3) The periodic noise signal is processed as follows:
the measurement of the periodic noise signal includes: carrying out standardization treatment on the source component matrix A to obtain a source component matrix C; carrying out peak statistics on each column of signals in the source component matrix C to obtain the peak value number of each column of signals; calculating the peak frequency p of each column of signals in unit time m2 The method comprises the steps of carrying out a first treatment on the surface of the If the peak frequency p m2 > set frequency threshold p 0 The corresponding column signal is marked as a periodic noise signal sub-component, and the position of the periodic noise signal sub-component is set as the position L of the periodic noise signal z . The measurement mode of the periodic noise signal is the same as the measurement mode of the time position of the low-frequency noise signal, and is not repeated here.
The separation of the periodic noise signal comprises: position L is matched by adopting a template matching method z On periodic interfering subcomponentsProcessing to separate out periodic noise signals. The periodic noise signal is for example an electrocardiographic noise signal. The template matching method comprises the following processing steps: detecting periodic interference subcomponents through a QRS wave detection algorithm, determining the positions of R waves in the subcomponents, extracting all QRS waves in signals according to the positions of the R waves, iteratively clustering all QRS wave signals to obtain a plurality of periodic noise signal templates, then matching the periodic interference subcomponents with the periodic noise signal templates one by one, recording the periodic noise signal template X with the highest matching degree, and subtracting the periodic noise signal template X from the subcomponents to obtain clean subcomponents.
The following is a comparative description by means of specific examples.
Example 1: the method is used for removing noise signals from simulation data, wherein the simulation data comprise electro-oculogram noise signals and myoelectric artifacts with different signal to noise ratios.
Comparative example 1: the simulation data is noise signal removed by adopting the existing MARA (multiple noise signal removal algorithm, multiple Artifact Rejection Algorithm) method.
Comparative example 2: the simulation data is noise signal removed by adopting the existing ASR (noise signal subspace reconstruction, artifact Subspace Reconstruction) method.
Example 2: the method is used for removing the noise signal from the real brain electrical data containing the electro-oculogram noise signal.
Comparative example 3: and removing noise signals from the real electroencephalogram data by adopting the existing MARA method.
Comparative example 4: and adopting the existing ASR method to remove noise signals from the real brain electrical data.
The evaluation indexes of the processing effect on the simulation data are as follows: the higher the correlation coefficient and the relative root mean square error between the data after removing the noise signal and the data without the noise signal, the better the artifact removing effect is, and the smaller the relative root mean square error is, the better the surface rolling noise signal effect is.
The evaluation indexes of the processing effect on the real data are as follows: waveform comparison between the original electroencephalogram data and the electroencephalogram data after noise signal removal.
Fig. 4 (a) is a graph showing correlation coefficients of the strong eye electric path of example 1, comparative example 1, and comparative example 2. Fig. 4 (b) is a graph showing correlation coefficients of the weak eye electric path of example 1, comparative example 1, and comparative example 2. Fig. 4 (c) is a graph showing correlation coefficients of the myoelectric channels of example 1, comparative example 1, and comparative example 2. Fig. 5 (a) is a graph of the relative root mean square error of example 1, comparative example 1, and comparative example 2 in the strong eye electrical channel. Fig. 5 (b) is a graph of the relative root mean square error of the weak eye electrical path for example 1, comparative example 1, and comparative example 2. Fig. 5 (c) is a graph of the relative root mean square error of the myoelectric channels for example 1, comparative example 1, and comparative example 2. The abscissa of the numeral "1" represents example 1, "2" represents comparative example 1, and "3" represents comparative example 2. From the results of fig. 4 and 5, the correlation coefficient of the method is higher than that of the prior art, and the relative root mean square error of the method is lower than that of the prior art. Therefore, when the method is used for removing the multiple types of noise signals, not only can the noise signals of different types be removed more accurately, but also the loss of effective signals can be reduced when the noise signals are removed.
Fig. 6 is a waveform comparison of the original brain electrical signal, example 2, comparative example 3, comparative example 4. The abscissa is time and the ordinate is amplitude. As can be seen from the figure, the original signal contains a plurality of peaks (i.e. noise signals), and the peaks are filtered out after processing by the method and two prior art techniques. Compared with the MARA method, the method has less loss of signal details and less influence on effective signals when noise signals are removed. Compared with the ASR method, the method has the advantages that partial noise signals still exist in the waveform processed by the ASR method, the removing effect is poor, and the waveform after the noise signals are removed by the method is more stable.
For example, the detection of the signal time window may employ a time window detection method, where the time window detection method includes: at least one of generalized characteristic decomposition method, multi-characteristic classification method and multi-characteristic joint detection method. The time window detection method is suitable for a primary signal time window or a secondary signal time window. When the secondary signal time window exists, the secondary signal time window is marked by a time window detection method, and then the category of the primary signal time window is judged in a combined mode through marking results of a plurality of secondary signal time windows, and the primary signal time window is divided into a sparse noise signal time window, a dense noise signal time window and a normal signal time window.
The generalized feature decomposition method as shown in fig. 8 includes: calculating a characteristic matrix of a signal time window and a background signal respectively; calculating the maximum generalized eigenvalue between the covariance matrix of the signal time window and the covariance matrix of the background signal; if the maximum generalized characteristic value is larger than the set threshold value, marking the signal time window as a noise signal time window; otherwise, the normal signal time window is marked. It should be noted that, the background signal is an electroencephalogram baseline or an electroencephalogram signal in a full frequency band, the generalized eigenvalue can reflect the difference degree between the two matrixes, and the class of the signal time window is identified through the magnitude of the generalized eigenvalue. For example, let R 1 Is the covariance matrix of the current signal time window, R 2 Is covariance matrix after high-pass filtering current signal time window, R b Is the covariance matrix of the background signal, R 2 And R is 1 Maximum generalized eigenvalue α between 1 Represents myoelectric noise signal intensity of current signal time window, R 1 And R is R b Maximum generalized eigenvalue alpha of (a) 2 The noise signal strength (without distinguishing the noise signal type) of the current signal time window. If alpha 1 If the signal time window is larger than the set threshold, the signal time window is marked as the myoelectric noise signal time window, otherwise, the signal time window is marked as the normal signal time window. If alpha is 2 If the signal time window is larger than the set threshold, the signal time window is marked as a noise signal time window, otherwise, the signal time window is marked as a normal signal time window.
For example, fig. 11 is a diagram of the effect of denoising signal processing after a noise signal is identified by a generalized characteristic decomposition method (noise signal is processed only in part), and fig. 12 is a diagram of the effect of denoising signal processing for an entire signal directly by using the prior art. As can be seen from comparing fig. 11 and 12, the noise signal is suppressed by the present method and the prior art, but the loss of the effective part of the signal is less by the present method than by the prior art.
As shown in fig. 9, the multi-feature joint detection method includes: calculating a plurality of characteristic values of the signal time window; and judging the threshold value of the plurality of characteristic values one by one, and marking the signal time window as a noise signal time window when the plurality of characteristic values of the signal time window meet the threshold value condition. For example, the characteristic value may be a line length, a zero-crossing rate, a mean value, or the like. And calculating K characteristic values of the signal time window, judging the K characteristic values one by one, wherein each characteristic value corresponds to a set threshold value, and judging that the signal time window is a noise signal time window when the K characteristic values meet the set threshold values, or else, judging that the signal time window is a normal signal time window. For example, a certain signal time window calculates two characteristic values of the line length and the zero crossing rate, when the line length characteristic meets a set threshold value, the signal time window is marked as an alternative interference, and if the zero crossing rate characteristic also meets the set threshold value, the signal time window is finally marked as a noise signal time window. Therefore, the accuracy of noise signal judgment can be improved, and erroneous judgment can be prevented.
As shown in FIG. 13, the graph is a comparison of detection results using a multi-feature combination detection method and a single feature detection method. The abscissa represents time window, and the ordinate represents channel, and the first plot in fig. 13 represents original signals in the order from top to bottom, and the areas (only from left) selected by the black boxes are myoelectric noise signals, normal brain electricity, epileptic brain electricity and myoelectric noise signals respectively; the second small graph shows the search result obtained by using the line length feature only, the third small graph shows the detection result obtained by combining the line length feature and the zero crossing rate feature, and white bright points in the second small graph and the third small graph are represented as positions for judging myoelectric noise signals. From this, it can be known that the myoelectric noise signal, the normal signal and the epileptic signal exist in the original signal, and a single feature detection method is adopted to determine the useful epileptic signal as the noise signal, which may cause inaccurate early warning of epileptic later. The method can accurately identify the myoelectric noise signals and does not misjudge the epileptic signals as the myoelectric signals.
As shown in fig. 11, the multi-feature classification method includes: and calculating a plurality of characteristic values of the signal time window, inputting the characteristic values into a classifier for classification, and outputting a noise signal time window and a normal signal time window. The characteristic value is, for example, a line length, a zero-crossing rate, an average value, etc., and after calculating the characteristic value of the signal time window, the characteristic value is input into a classifier, and the classifier can output a classification result (noise signal or normal signal). And classifying the data of the M channels in the signal time window one by one to obtain a space-time domain classification result of the signal time window.
As shown in fig. 7, the noise signal class in the noise signal time window is measured further includes: dividing a noise signal time window into the following steps according to the space-time domain sparseness degree of the noise signal: sparse noise signal time window and dense noise signal time window. The divided signal time window comprises at least two stages of division, namely a first stage signal time window is formed by a plurality of second stage signal time windows. The classifying detection of the signal time window further comprises: and judging the primary signal time window based on the detection result of the secondary signal time window in a combined mode so as to divide the primary signal time window into a sparse noise signal time window, a dense noise signal time window and a normal signal time window. The step of jointly judging the primary signal time window based on the detection result of the secondary signal time window comprises the following steps: detecting a secondary signal time window, which is divided into a secondary interference time window and a secondary normal time window; calculating the duty ratio p of a second-level interference time window in the first-level signal time window, namely when the duty ratio p=0, marking the first-level signal time window as a normal signal time window; when the duty ratio 0<p is less than the duty ratio threshold, marking the first-level signal time window as a sparse noise signal time window; when the duty ratio p is more than or equal to the duty ratio threshold value, the primary signal time window is marked as a dense noise signal time window.
In other words, the present invention can classify noise signal time windows according to the degree of sparseness of noise signals, in addition to classifying noise signals according to the type of noise signals. The signal time window is divided into two stages, namely a second-stage signal time window and a first-stage signal time window, wherein the first-stage signal time window consists of a plurality of second-stage signal time windows, namely, the window width of the first-stage signal time window is larger than that of the second-stage signal time window, and the data volume contained in the first-stage signal time window is larger than that of the second-stage signal time window. Therefore, when the data processing is performed, the method and the device detect the secondary signal time windows to obtain the detection result (divided into the secondary interference time windows and the secondary normal time windows) of each secondary signal time window, so that the number of the secondary interference time windows contained in the primary signal time window can be reflected, and the number of the secondary interference time windows/the number of the secondary signal time windows with the duty ratio p=can be reflected. If p=0, this indicates no secondary interference window, the primary signal window is a normal signal window, and no further processing is required. If 0<p < duty ratio threshold value, it indicates that there is a small amount of second-level interference time window, and the first-level signal time window is a sparse noise signal time window, at this time, only the second-level interference time window needs to be processed when removing the noise signal, and the whole first-level signal time window does not need to be processed. If the duty ratio p is more than or equal to the duty ratio threshold value, the existing second-level interference time windows are more, the first-level signal time window is a dense noise signal time window, and at the moment, when noise signals are removed, the whole section of the first-level signal time window needs to be processed.
When the primary signal time window is a sparse noise signal time window, different signal processing models are built only for the secondary interference time window in the primary signal time window. When the primary signal time window is a dense noise signal time window, different signal processing models are constructed for the primary signal time window. Namely, the method only processes the second-level interference time window aiming at the first-level signal time window of sparse noise signal, so that the processing of a non-noise signal section can be avoided, and the loss of an effective signal is prevented; but also can reduce the data volume to be processed and improve the processing efficiency. Aiming at the first-level signal time window of dense noise signals, the whole section of the first-level signal time window is adopted for processing, so that the operation efficiency is higher, and the missing processing can be prevented.
It should be noted that the set threshold values in the present disclosure may be set according to requirements or historical data or technical experience, and are not specifically limited in the present disclosure.
According to the signal noise detection method, the time-space domain combined method is adopted to measure and separate the signals of multiple types of noise signals, so that the accuracy of noise signal removal can be improved, and the loss of effective signals is reduced. And for different types of noise signals, different measuring modes and separating modes are adopted, so that compared with the same processing mode, the method has more pertinence, and the identification result is more accurate.
The invention also provides a signal detection model of the signal noise detection method, which comprises the following steps: the signal time window dividing model is used for dividing the signal time window; the signal time window detection model is used for carrying out classification detection on the signal time windows to obtain noise signal time windows and normal signal time windows; a classification model for measuring noise signal categories in a noise signal time window; and the signal processing model separates noise signals of different categories according to the noise signal category. For the description of the relevant parts of the signal noise detection method, please refer to the parts of the signal noise detection method, and the description thereof is omitted herein.
The invention also provides a computer readable storage medium, on which a computer program is stored, which when being executed by a processor implements the method for detecting signal noise. The computer readable storage medium may be located in at least one of a plurality of network servers of a computer network. The storage medium may include, but is not limited to: a usb disk, a read-only memory (ROM), a random-access memory (RAM), a removable hard disk, a magnetic disk, or an optical disk, etc., which can store program codes.
The invention also provides a signal time window detection model based on the generalized characteristic decomposition method, which comprises the following steps: calculating a characteristic matrix of a signal time window and a background signal respectively; calculating the maximum generalized eigenvalue between the eigenvalue of the signal time window and the eigenvalue of the background signal; if the maximum generalized characteristic value is larger than the set threshold value, marking the signal time window as a noise signal time window; otherwise, the normal signal time window is marked. Details of the detection method of signal noise are described in detail, and are not repeated here.
The invention also provides a signal time window detection model based on the multi-feature classification method, which comprises the following steps: calculating a plurality of characteristic values of the signal time window; and judging the threshold value of the plurality of characteristic values one by one, and marking the signal time window as a noise signal time window when the plurality of characteristic values of the signal time window meet the threshold value condition. Details of the detection method of signal noise are described in detail, and are not repeated here.
The invention also provides a signal time window detection model based on the multi-feature joint detection method, which comprises the following steps: and calculating a plurality of characteristic values of the signal time window, inputting the characteristic values into a classifier for classification, and outputting a noise signal time window and a normal signal time window. Details of the detection method of signal noise are described in detail, and are not repeated here.
The invention also provides a high-frequency noise signal detection model, which comprises the following steps: dividing a signal time window; classifying and detecting the signal time window, wherein the signal time window is divided into a noise signal time window and a normal signal time window; a signal processing model is built for the noise signal time window to separate the high frequency noise signal. The signal processing model includes: acquiring a time decomposition matrix V based on a noise signal time window, wherein the time decomposition matrix V is a K multiplied by K matrix, and K is the number of time domain sampling points; extracting the first M columns from the time decomposition matrix V to form a source component matrix A, wherein M is the channel number; positioning the position of the high-frequency noise signal in the source component matrix A; judging whether the position of the high-frequency noise signal in the source component matrix A is a null value or not; when the position of the high-frequency noise signal is a non-null value, sequentially carrying out separation of the high-frequency noise signal and source reconstruction so as to continue iteration; when the position of the high frequency noise signal is null, the iteration is terminated. Locating the position of the high frequency noise signal in the source component matrix a includes: calculating the energy characteristics of each column of signals in the source component matrix A; calculating the high-low frequency energy ratio r of each column of signals hl The method comprises the steps of carrying out a first treatment on the surface of the If the high-low frequency energy ratio r hl > set energy ratio threshold r 0 Marking the corresponding column signal as high-frequency noise signal subcomponent, and setting the set of all high-frequency noise signal subcomponent positions as the position L of the high-frequency noise signal h . Separating the high frequency noise signal includes: to position L h The high-frequency interference subcomponent is subjected to low-pass filtering processing or CCA filtering processing to separate out a high-frequency noise signal. Details of the detection method of signal noise are described in detail, and are not repeated here.
The invention also provides a low-frequency noise signal detection model, which comprises the following steps: dividing a signal time window; classifying and detecting the signal time window, wherein the signal time window is divided into a noise signal time window and a normal signal time window; a signal processing model is built for the noise signal time window to separate the low frequency noise signal. The signal processing model includes: acquiring a space decomposition matrix U and a time decomposition matrix V based on a noise signal time window, wherein the space decomposition matrix U is an M×M matrix, M is the channel number, the time decomposition matrix V is a K×K matrix, and K is timeDomain sampling points; extracting the first M columns from the time decomposition matrix V to form a source component matrix A, wherein M is the channel number; positioning the position of a low-frequency noise signal in the source component matrix A; judging whether the position of the low-frequency noise signal in the source component matrix A is a null value or not; when the position of the low-frequency noise signal is a non-null value, sequentially carrying out separation of the low-frequency noise signal and source reconstruction so as to continue iteration; when the position of the low frequency noise signal is null, the iteration is terminated. Locating the position of the low frequency noise signal in the source component matrix a includes: carrying out standardization treatment on the source component matrix A to obtain a source component matrix B; carrying out peak statistics on each column of signals in the source component matrix B to obtain the peak value number of each column of signals; calculating the peak frequency p of each column of signals in unit time m1 The method comprises the steps of carrying out a first treatment on the surface of the If the peak frequency p m1 > set frequency threshold p 0 The corresponding column signal is marked as a low-frequency noise signal sub-component, and the time position L of the low-frequency noise signal sub-component is recorded v The method comprises the steps of carrying out a first treatment on the surface of the Matching the mapped residual space position templates of each column of signals of the space decomposition matrix U to obtain a matching coefficient; if the matching coefficient is greater than the set matching threshold, outputting the spatial position L of the low-frequency noise signal subcomponent u The method comprises the steps of carrying out a first treatment on the surface of the According to formula L l =merge(L v ,L u ) Determining the position L of a low-frequency noise signal l . Separating the low frequency noise signal includes: to position L l The low-frequency interference subcomponent is subjected to wavelet threshold filtering processing to separate out a low-frequency noise signal. Details of the detection method of signal noise are described in detail, and are not repeated here.
The invention also provides a periodic noise signal detection model, comprising: dividing a signal time window; classifying and detecting the signal time window, wherein the signal time window is divided into a noise signal time window and a normal signal time window; a signal processing model is constructed for the noise signal time window to separate the periodic noise signal. The signal processing model includes: acquiring a time decomposition matrix V based on a noise signal time window, wherein the time decomposition matrix V is a K multiplied by K matrix, and K is the number of time domain sampling points; extracting the first M columns from the time decomposition matrix V to form a source component matrix A, wherein M is the channel number; positioning the position of the periodic noise signal in the source component matrix A; judging whether the source is finished Whether the position of the periodic noise signal in the subarray A is a null value or not; when the position of the periodic noise signal is a non-null value, sequentially carrying out separation periodic noise signal and source reconstruction so as to continue iteration; when the position of the periodic noise signal is null, the iteration is terminated. Locating the position of the periodic noise signal in the source component matrix a includes: carrying out standardization treatment on the source component matrix A to obtain a source component matrix C; carrying out peak statistics on each column of signals in the source component matrix C to obtain the peak value number of each column of signals; calculating the peak frequency p of each column of signals in unit time m2 The method comprises the steps of carrying out a first treatment on the surface of the If the peak frequency p m2 > set frequency threshold p 0 The corresponding column signal is marked as a periodic noise signal sub-component, and the position of the periodic noise signal sub-component is set as the position L of the periodic noise signal z . The separation of the periodic noise signal comprises: position L is matched by adopting a template matching method z The periodic interfering subcomponents are processed to separate out a periodic noise signal. Details of the detection method of signal noise are described in detail, and are not repeated here.
The invention also provides a multi-stage detection model of the signal time window, which comprises the following steps: at least two stages of dividing signal time windows, namely forming a first-stage signal time window by a plurality of second-stage signal time windows; detecting a secondary signal time window by using a time window detection method, wherein the secondary signal time window is divided into a secondary interference time window and a secondary normal time window; and judging the primary signal time window based on the detection result of the secondary signal time window in a combined mode so as to divide the primary signal time window into a sparse noise signal time window, a dense noise signal time window and a normal signal time window. The time window detection method is at least one of a generalized characteristic decomposition algorithm, a multi-characteristic classification algorithm and a multi-characteristic joint detection algorithm. The generalized feature decomposition method includes: calculating a characteristic matrix of a signal time window and a background signal respectively; calculating the maximum generalized eigenvalue between the eigenvalue of the signal time window and the eigenvalue of the background signal; if the maximum generalized characteristic value is larger than the set threshold value, marking the signal time window as a noise signal time window; otherwise, the normal signal time window is marked. The multi-feature joint detection method comprises the following steps: calculating a plurality of characteristic values of the signal time window; and judging the threshold value of the plurality of characteristic values one by one, and marking the signal time window as a noise signal time window when the plurality of characteristic values of the signal time window meet the threshold value condition. The multi-feature classification method comprises the following steps: and calculating a plurality of characteristic values of the signal time window, inputting the characteristic values into a classifier for classification, and outputting a noise signal time window and a normal signal time window. The step of jointly judging the primary signal time window based on the detection result of the secondary signal time window comprises the following steps: calculating the duty ratio p of a second-level interference time window in the first-level signal time window; when the duty ratio p=0, marking the first-level signal time window as a normal signal time window; when the duty ratio 0<p is less than the duty ratio threshold, marking the first-level signal time window as a sparse noise signal time window; when the duty ratio p is more than or equal to the duty ratio threshold value, the primary signal time window is marked as a dense noise signal time window. Details of the detection method of signal noise are described in detail, and are not repeated here.
In summary, the signal noise detection method, the signal detection model and the readable storage medium of the invention adopt the time-space domain combined method to measure and separate the signals of multiple types of noise signals, can improve the accuracy of noise signal removal and reduce the loss of effective signals. And for different types of noise signals, different measuring modes and separating modes are adopted, so that compared with the same processing mode, the method has more pertinence, and the identification result is more accurate.
With the above-described preferred embodiments according to the present invention as an illustration, the above-described descriptions can be used by persons skilled in the relevant art to make various changes and modifications without departing from the scope of the technical idea of the present invention. The technical scope of the present invention is not limited to the description, but must be determined as the scope of the claims.
Claims (26)
1. A method for detecting signal noise, comprising:
dividing a signal time window;
classifying and detecting the signal time window, wherein the signal time window is divided into a noise signal time window and a normal signal time window;
detecting a noise signal class in the noise signal time window;
constructing different signal processing models for the noise signal time window according to the noise signal category so as to separate out corresponding noise signals;
Wherein detecting the noise signal class in the noise signal time window comprises:
detecting a non-biological source noise signal and a biological source noise signal in the noise signal; and/or detecting a noise signal category according to the space-time domain sparseness degree of the noise signal;
the non-biological source noise signal comprises: power frequency noise, environmental electromagnetic interference, step noise;
the biological source noise signal comprises: at least one of a high frequency noise signal, a low frequency noise signal, and a periodic noise signal;
the signal processing model includes:
acquiring a time decomposition matrix V and a space decomposition matrix U based on the noise signal time window, wherein the time decomposition matrix V is a K multiplied by K matrix, and K is the number of time domain sampling points; the space decomposition matrix U is an M×M matrix, and M is the channel number;
extracting the first M columns from the time decomposition matrix V to form a source component matrix A, wherein M is the channel number;
locating the position of the noise signal in the source component matrix a comprises:
positioning the position of power frequency noise and environmental electromagnetic noise signals through the space characteristics and the time-frequency characteristics;
performing first-order difference processing on each column of the source component matrix A, and determining the position of a step noise signal according to the absolute value of a difference value;
Positioning the position of a high-frequency noise signal according to the characteristics of each column of signals in the source component matrix A;
carrying out standardization processing on the source component matrix A and positioning the positions of a low-frequency noise signal and a periodic noise signal;
judging whether the position of a noise signal in the source component matrix A is a null value, and when the position of the noise signal is a non-null value, sequentially carrying out noise signal separation and source reconstruction so as to continue iteration; terminating the iteration when the position of the noise signal is null;
the source reconstruction includes:
outputting a matrix V_hat after filtering the noise signals;
and performing source decomposition inverse operation on the space decomposition matrix U and the matrix V_hat to obtain a signal S_hat.
2. The method for detecting signal noise according to claim 1, wherein,
setting the priority order of the non-biological source noise signals and biological source noise signals;
and sequentially operating different signal processing models according to the priority levels so as to separate noise signals with different levels step by step.
3. The method for detecting signal noise according to claim 2, wherein,
setting the priority order between the non-biological source noise signals and the biological source noise signals according to the energy of the noise signals;
The priority order between the non-biological source noise signals and the biological source noise signals is as follows: non-biogenic noise signal > biogenic noise signal.
4. The method for detecting signal noise according to claim 1, wherein,
the priority order of the non-biological source noise signals is as follows: the power frequency noise is greater than the environmental electromagnetic interference and the step noise.
5. The method for detecting signal noise according to claim 1, wherein,
the separation of the power frequency noise and the environmental electromagnetic interference comprises the following steps:
and filtering the source components containing the power frequency noise and the environmental electromagnetic interference to separate the power frequency noise and the environmental electromagnetic noise signals.
6. The method for detecting signal noise according to claim 1, wherein,
the step noise separation includes:
and performing wavelet threshold filtering processing on the components corresponding to the step noise signals to separate out the step noise signals.
7. The method for detecting signal noise according to claim 1, wherein,
the priority order of the biological source noise signals is as follows: high frequency noise signal > low frequency noise signal > periodic noise signal.
8. The method for detecting signal noise according to claim 1, wherein,
The measurement of the high frequency noise signal includes:
calculating the energy characteristics of each column of signals in the source component matrix A;
calculating the high-low frequency energy ratio r of each column of signals hl ；
If the high-low frequency energy ratio r hl > set energy ratio threshold r 0 Marking the corresponding column signal as high-frequency noise signal subcomponent, and setting the set of all high-frequency noise signal subcomponent positions as the position L of the high-frequency noise signal h 。
9. The method for detecting signal noise according to claim 8, wherein,
the separation of the high frequency noise signal comprises:
to position L h The high-frequency interference subcomponent is subjected to filtering processing or CCA filtering processing to separate out a high-frequency noise signal.
10. The method for detecting signal noise according to claim 1, wherein,
the measurement of the low frequency noise signal comprises:
performing standardization processing on the source component matrix A to obtain a source component matrix B;
carrying out peak statistics on each column of signals in the source component matrix B to obtain the peak value number of each column of signals;
calculating the peak frequency p of each column of signals in unit time m1 ；
If the peak frequency p m1 > set frequency threshold p 0 The corresponding column signal is marked as a low-frequency noise signal sub-component, and the time position L of the low-frequency noise signal sub-component is recorded v ；
Acquiring a space decomposition matrix U based on the noise signal time window, wherein the space decomposition matrix U is an M multiplied by M matrix, and M is the channel number;
each column of signals of the space decomposition matrix U is mapped and then matched with a space position template to obtain a matching coefficient;
if the matching coefficient is greater than the set matching threshold, outputting the spatial position L of the low-frequency noise signal subcomponent u ；
According to formula L l =merge(L v ,L u ) Determining the position L of a low-frequency noise signal l 。
11. The method for detecting signal noise according to claim 10, wherein,
the separation of the low frequency noise signal comprises:
to position L l The low-frequency noise signal subcomponents are subjected to wavelet threshold filtering processing to separate out the low-frequency noise signal.
12. The method for detecting signal noise according to claim 1, wherein,
the measurement of the periodic noise signal comprises:
performing standardization processing on the source component matrix A to obtain a source component matrix C;
carrying out peak statistics on each column of signals in the source component matrix C to obtain the peak value number of each column of signals;
calculating the peak frequency p of each column of signals in unit time m2 ；
If the peak frequency p m2 > set frequency threshold p 0 The corresponding column signal is marked as a periodic noise signal sub-component, and the position of the periodic noise signal sub-component is set as the position L of the periodic noise signal z 。
13. The method for detecting signal noise according to claim 12, wherein,
the separation of the periodic noise signals includes:
the position L is matched by a template matching method z The periodic interfering subcomponents are processed to separate out a periodic noise signal.
14. The method for detecting signal noise according to claim 1, wherein,
detecting the signal time window by using a time window detection method;
the time window detection method comprises the following steps: at least one of generalized characteristic decomposition method, multi-characteristic classification method and multi-characteristic joint detection method.
15. The method for detecting signal noise according to claim 14, wherein,
the generalized characteristic decomposition method includes: respectively calculating covariance matrixes of the signal time windows and the background signals;
calculating the maximum generalized eigenvalue between the covariance matrix of the signal time window and the covariance matrix of the background signal;
if the maximum generalized characteristic value is larger than the set threshold value, marking the signal time window as a noise signal time window; otherwise, the normal signal time window is marked.
16. The method for detecting signal noise according to claim 14, wherein,
the multi-feature joint detection method comprises the following steps:
Calculating a plurality of characteristic values of the signal time window;
and judging the threshold value of the plurality of characteristic values one by one, and marking the signal time window as a noise signal time window when the plurality of characteristic values of the signal time window meet the threshold value condition.
17. The method for detecting signal noise according to claim 14, wherein,
the multi-feature taxonomy comprises:
and calculating a plurality of characteristic values of the signal time window, inputting the characteristic values into a classifier for classification, and outputting a noise signal time window and a normal signal time window.
18. The method for detecting signal noise according to claim 1, wherein,
dividing the noise signal time window into the following steps according to the space-time domain sparseness degree of the noise signal: sparse noise signal time window and dense noise signal time window.
19. The method for detecting signal noise according to claim 18, wherein,
the dividing signal time window comprises at least two stages of division, namely a first-stage signal time window is formed by a plurality of second-stage signal time windows;
the classifying detection of the signal time window further comprises:
and judging a primary signal time window based on the detection result of the secondary signal time window in a combined mode so as to divide the primary signal time window into a sparse noise signal time window, a dense noise signal time window and a normal signal time window.
20. The method for detecting signal noise according to claim 19, wherein,
the detection result of the secondary signal time window jointly judges the primary signal time window comprises:
detecting a secondary signal time window, which is divided into a secondary interference time window and a secondary normal time window;
calculating the duty ratio p of the second-level interference time window in the first-level signal time window, namely
When the duty ratio p=0, marking the primary signal time window as a normal signal time window;
marking the primary signal time window as a sparse noise signal time window when the duty cycle 0<p < duty cycle threshold;
and when the duty ratio p is more than or equal to the duty ratio threshold value, marking the primary signal time window as a dense noise signal time window.
21. The method for detecting signal noise according to claim 20, wherein,
when the primary signal time window is a sparse noise signal time window, constructing different signal processing models for the secondary interference time window in the primary signal time window only;
when the primary signal time window is a dense noise signal time window, different signal processing models are constructed for the primary signal time window.
22. A signal detection apparatus that operates the signal noise detection method according to any one of claims 1 to 21, comprising:
the signal time window dividing module is used for dividing the signal time window;
The signal time window detection module is used for carrying out classified detection on the signal time windows to obtain noise signal time windows and normal signal time windows;
the classification module is used for detecting the noise signal category in the noise signal time window;
and the signal processing module separates noise signals of different categories according to the noise signal category.
23. A computer readable storage medium, wherein a computer program is stored on the computer readable storage medium, which when executed by a processor implements the method of detecting signal noise according to any one of claims 1 to 21.
24. A signal time window detection device based on a generalized characteristic decomposition method, comprising:
covariance matrix of signal time window and covariance matrix of background signal, and maximum generalized eigenvalue between them and its set threshold; wherein the method comprises the steps of
The signal time window detecting apparatus operates the signal noise detecting method as claimed in claim 15 to divide the signal time window into a noise signal time window and a normal signal time window according to a generalized characteristic decomposition method.
25. A signal time window detection device based on a multi-feature joint detection method, comprising:
A plurality of characteristic values of the signal time window and corresponding threshold conditions; wherein the method comprises the steps of
The signal time window detection device operates the signal noise detection method according to claim 16 to divide the signal time window into a noise signal time window and a normal signal time window according to the multi-feature joint detection method.
26. A signal time window detection device based on a multi-feature classification method, comprising:
a plurality of characteristic values of the classifier and the signal time window; wherein the method comprises the steps of
The signal time window detection device operates the signal noise detection method as claimed in claim 17, so as to divide the signal time window into a noise signal time window and a normal signal time window according to the multi-feature classification method.
Priority Applications (5)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311044638.0A CN117034087A (en) | 2023-06-06 | 2023-06-06 | Multistage detection model of signal time window |
| CN202311040706.6A CN117034085A (en) | 2023-06-06 | 2023-06-06 | Periodic noise signal detection model |
| CN202310658542.7A CN116383709B (en) | 2023-06-06 | 2023-06-06 | Signal noise detection method, signal detection model, and readable storage medium |
| CN202311040552.0A CN117034084A (en) | 2023-06-06 | 2023-06-06 | High-frequency noise signal detection model |
| CN202311040644.9A CN117093897A (en) | 2023-06-06 | 2023-06-06 | Low-frequency noise signal detection model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310658542.7A CN116383709B (en) | 2023-06-06 | 2023-06-06 | Signal noise detection method, signal detection model, and readable storage medium |
Related Child Applications (4)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311044638.0A Division CN117034087A (en) | 2023-06-06 | 2023-06-06 | Multistage detection model of signal time window |
| CN202311040552.0A Division CN117034084A (en) | 2023-06-06 | 2023-06-06 | High-frequency noise signal detection model |
| CN202311040644.9A Division CN117093897A (en) | 2023-06-06 | 2023-06-06 | Low-frequency noise signal detection model |
| CN202311040706.6A Division CN117034085A (en) | 2023-06-06 | 2023-06-06 | Periodic noise signal detection model |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116383709A CN116383709A (en) | 2023-07-04 |
| CN116383709B true CN116383709B (en) | 2023-09-15 |
Family
ID=86961930
Family Applications (5)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311040552.0A Pending CN117034084A (en) | 2023-06-06 | 2023-06-06 | High-frequency noise signal detection model |
| CN202311044638.0A Pending CN117034087A (en) | 2023-06-06 | 2023-06-06 | Multistage detection model of signal time window |
| CN202311040644.9A Pending CN117093897A (en) | 2023-06-06 | 2023-06-06 | Low-frequency noise signal detection model |
| CN202311040706.6A Pending CN117034085A (en) | 2023-06-06 | 2023-06-06 | Periodic noise signal detection model |
| CN202310658542.7A Active CN116383709B (en) | 2023-06-06 | 2023-06-06 | Signal noise detection method, signal detection model, and readable storage medium |
Family Applications Before (4)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311040552.0A Pending CN117034084A (en) | 2023-06-06 | 2023-06-06 | High-frequency noise signal detection model |
| CN202311044638.0A Pending CN117034087A (en) | 2023-06-06 | 2023-06-06 | Multistage detection model of signal time window |
| CN202311040644.9A Pending CN117093897A (en) | 2023-06-06 | 2023-06-06 | Low-frequency noise signal detection model |
| CN202311040706.6A Pending CN117034085A (en) | 2023-06-06 | 2023-06-06 | Periodic noise signal detection model |
Country Status (1)
| Country | Link |
|---|---|
| CN (5) | CN117034084A (en) |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106663446A (en) * | 2014-07-02 | 2017-05-10 | 微软技术许可有限责任公司 | User environment aware acoustic noise reduction |
| CN114330459A (en) * | 2022-01-06 | 2022-04-12 | 北京交通大学 | Vibration source response signal-noise separation method and system based on multi-layer decomposition |
-
2023
- 2023-06-06 CN CN202311040552.0A patent/CN117034084A/en active Pending
- 2023-06-06 CN CN202311044638.0A patent/CN117034087A/en active Pending
- 2023-06-06 CN CN202311040644.9A patent/CN117093897A/en active Pending
- 2023-06-06 CN CN202311040706.6A patent/CN117034085A/en active Pending
- 2023-06-06 CN CN202310658542.7A patent/CN116383709B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106663446A (en) * | 2014-07-02 | 2017-05-10 | 微软技术许可有限责任公司 | User environment aware acoustic noise reduction |
| CN114330459A (en) * | 2022-01-06 | 2022-04-12 | 北京交通大学 | Vibration source response signal-noise separation method and system based on multi-layer decomposition |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117093897A (en) | 2023-11-21 |
| CN116383709A (en) | 2023-07-04 |
| CN117034084A (en) | 2023-11-10 |
| CN117034085A (en) | 2023-11-10 |
| CN117034087A (en) | 2023-11-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107844755B (en) | Electroencephalogram characteristic extraction and classification method combining DAE and CNN | |
| Safara et al. | Multi-level basis selection of wavelet packet decomposition tree for heart sound classification | |
| Prochazka et al. | Wavelet transform use for feature extraction and EEG signal segments classification | |
| Chawla | A comparative analysis of principal component and independent component techniques for electrocardiograms | |
| US20070260151A1 (en) | Method and device for filtering, segmenting, compressing and classifying oscillatory signals | |
| CN112957055B (en) | EEMD-PCA-based method and device for removing motion artifacts in EEG signals | |
| US9125580B2 (en) | Electrocardiography signal extraction method | |
| Chashmi et al. | An efficient and automatic ECG arrhythmia diagnosis system using DWT and HOS features and entropy-based feature selection procedure | |
| CN113066502A (en) | Heart sound segmentation positioning method based on VMD and multi-wavelet | |
| Sharathappriyaa et al. | Auto-encoder based automated epilepsy diagnosis | |
| CN116211322A (en) | Depression recognition method and system based on machine learning electroencephalogram signals | |
| Nguyen et al. | Artifact elimination in ECG signal using wavelet transform | |
| CN115486849A (en) | Electrocardiosignal quality evaluation method, device and equipment | |
| CN116383709B (en) | Signal noise detection method, signal detection model, and readable storage medium | |
| CN112033656A (en) | Mechanical system fault detection method based on broadband spectrum processing | |
| Saxena et al. | Optimisation of thresholding techniques in de-noising of ECG signals | |
| CN115067878A (en) | EEGNet-based resting state electroencephalogram consciousness disorder classification method and system | |
| CN114947850A (en) | Mental load grade objective detection method based on pulse Bouss model characteristics | |
| Banerjee et al. | A classification approach for myocardial infarction using voltage features extracted from four standard ECG leads | |
| Islam et al. | Wavelet-based Artifact Removal Algorithm for EEG Data by Optimizing Mother Wavelet and Threshold Parameters | |
| CN112378942A (en) | White spirit grade classification identification method based on nuclear magnetic resonance fingerprint spectrum | |
| NSVN et al. | Optimal threshold estimation using cultural algorithm for EMD-DWT based ECG denoising | |
| Ince et al. | ECoG based brain computer interface with subset selection | |
| Mohammed et al. | Detecting signatures in hyperspectral image data of wounds: a compound model of self-organizing map and least square fitting | |
| Thatipelli et al. | Electrocardiogram beat classification using discrete wavelet transform, higher order statistics and multivariate analysis |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |