CN102217932B - Brand-new algorithm for ABR (auditory brainstem response) signal crest detection - Google Patents

Brand-new algorithm for ABR (auditory brainstem response) signal crest detection Download PDF

Info

Publication number
CN102217932B
CN102217932B CN 201110126653 CN201110126653A CN102217932B CN 102217932 B CN102217932 B CN 102217932B CN 201110126653 CN201110126653 CN 201110126653 CN 201110126653 A CN201110126653 A CN 201110126653A CN 102217932 B CN102217932 B CN 102217932B
Authority
CN
China
Prior art keywords
abr
window
signal
sum
point
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.)
Expired - Fee Related
Application number
CN 201110126653
Other languages
Chinese (zh)
Other versions
CN102217932A (en
Inventor
高瑞静
孙迎
李越囡
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
University of Shanghai for Science and Technology
Original Assignee
University of Shanghai for Science and Technology
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by University of Shanghai for Science and Technology filed Critical University of Shanghai for Science and Technology
Priority to CN 201110126653 priority Critical patent/CN102217932B/en
Publication of CN102217932A publication Critical patent/CN102217932A/en
Application granted granted Critical
Publication of CN102217932B publication Critical patent/CN102217932B/en
Expired - Fee Related legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Landscapes

  • Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)

Abstract

The invention relates to a brand-new algorithm for ABR (auditory brainstem response) signal crest detection. A superposition average technology is adopted to denoise an acquired original ABR signal, then automatic identification is carried out on crest detection of an ABR wave, no matter whether the ABR signal is a sensation level signal or a hearing level signal, all the crests in a wave form can be easily and accurately detected by adopting the method, each ABR subwave can be independently analyzed, and effective ABR parameters are provided for hearing threshold evaluation and surgical operation monitoring.

Description

The algorithm that a kind of ABR signal wave crest detects
Technical field
The present invention relates to a kind of detection technique, particularly a kind of algorithm of ABR signal wave crest detection.
Background technology
Brainstem auditory evoked potential ABR uses the most widely one of auditory evoked potential clinically.ABR is the bioelectric of Auditory neuropathway before brain stem and the brain stem, and it is to stimulate afterwards and induce a series of current potentials incubation period in 10ms lack sound, belongs to short delitescence and brings out current potential.Can be recorded to 7 positive phase waves the adult, usually use the Roman number I, II, III, IV, V, VI, VII represent, and be wherein the most obvious with I, III, V ripple.These ripples are with the degree that incubation period of determining and uncertain amplitude embody the main body threshold of audibility, hearing loss, perhaps are used for real-time monitored auditory nerve situation.In the threshold of audibility evaluate application, we need Automatic Logos ABR waveform, for clinical practice provides assistance in diagnosis.
At present, the method that detects the positive crest of ABR has: the wave filter that the local maximum, design of directly searching ABR ripple after the filtering and interested crest mate, utilize neutral net, also have the most frequently used method to be: the wavestrip of restriction ripple, search the zero crossing of waveform first derivative, the different Wave crest and wave troughs of distinguishing of direction according to zero crossing have namely found crest.In addition, also proposed many decomposition detection method, at first narrow frequency band filter detects crest, then obtains more accurate crest time Estimate with a wider bandpass filters.Because narrow band filter has the signal to noise ratio (SNR) of an improvement, simultaneously, the broad frequency band wave filter has improved the definition of crest time, has therefore thisly really improved detection under the low stimulus intensity by coarse to accurate method.Yet, these detection methods mainly are applicable to the estimation of threshold of audibility, and be merely able to detect limited crest, as detecting amplitude and incubation period more stable I all, III, the V ripple, perhaps sometimes detecting the V ripple is the stack ripple of original IV and V ripple, this can bring to diagnosis clinically very big inconvenience unavoidably.
Summary of the invention
The present invention be directed to the circumscribed problem of system detection method, proposed the algorithm that a kind of ABR signal wave crest detects, the crest that is applied to the ABR ripple based on the parser of vector detects and Automatic Logos, realizes simply, has high accuracy.
Technical scheme of the present invention is: the algorithm that a kind of ABR signal wave crest detects, and use vector operation that the ABR crest is carried out Automatic Logos, comprise following concrete steps:
1) denoising of original ABR signal:
Adopt the superposed average technology that the original ABR signal that collects is carried out denoising, namely apply N stimulation, the discretization data were adjusted into the moment of each stimulation appearance align, then average.Be that these adopt ABR signal that superposed average methods draw all based on following hypothesis:
A) under identical stimulation, the waveform of brain evoked response signal is constant, and namely response signal is deterministic signal;
B) background noise has randomness, through repeatedly superposeing and can eliminating.And, the pass of it and ABR signal is f (t)=s (t)+n (t), s in the formula (t) is the real not ABR signal of Noise, n (t) is background noise, f (t) is the measurement data of ABR signal+noise, under such hypothesis, obtained relatively level and smooth ABR signal;
2) the ABR ripple is 2D signal, establishes the sampling time to be x, sample magnitude is y, namely sampled point be expressed as ( x, y), for the signal curve that collects, get window area, be made as A, value is respectively y 1 , y 2.。。, y m ,, the sampling time is respectively x 1 , x 2 ..., x m , altogether get mIt is individual, mBe odd number, the coordinate of each point be ( x n , y n ), wherein n=1,2... m, then the middle point coordinates in the window area is
Figure DEST_PATH_IMAGE001
, the vector between other points of window area and the mid point can be expressed as
Figure 2011101266530100002DEST_PATH_IMAGE002
, the difference of the amplitude between 2 is
Figure DEST_PATH_IMAGE003
, existing along y direction of principal axis amount of orientation (0 ,-△ y), be designated as
Figure 2011101266530100002DEST_PATH_IMAGE004
, compute vector then
Figure DEST_PATH_IMAGE005
With
Figure 965603DEST_PATH_IMAGE004
The cosine value of angle is:
Figure 2011101266530100002DEST_PATH_IMAGE006
, vector between each point and the mid point in the calculation window zone
Figure 820427DEST_PATH_IMAGE005
And vector
Figure 285650DEST_PATH_IMAGE004
The included angle cosine value
Figure DEST_PATH_IMAGE007
, and among the accumulative total window A
Figure 618542DEST_PATH_IMAGE007
Value is calculated SUM=
Figure 2011101266530100002DEST_PATH_IMAGE008
3) establish total N the data point of ABR data, window size is m point, and take step-length as 1 moving window, window size is constant, and the window number that then obtains is N-m+1, is designated as A 1, A 2A N-m+1Accumulate successively in each window
Figure 619865DEST_PATH_IMAGE007
Value calculates corresponding SUM 1, SUM 2SUM N-m+1, and record the mid point coordinate figure of each window; By the characteristic of cosine function as can be known,
Figure DEST_PATH_IMAGE009
With
Figure 2011101266530100002DEST_PATH_IMAGE010
Cosine value is being for just,
Figure DEST_PATH_IMAGE011
Cosine value is worked as SUM for negative nSUM N-1, and SUM nSUM N+1, corresponding window A then nThe interior maximum that occurred, maximum are window A nMid point;
4) with the cosine value of each window and SUM by from big to small ordering, come the larger window of front SUM and just comprised all extreme points; In these points, select the crest of ABR ripple I, II, III, IV, V.
Beneficial effect of the present invention is: the algorithm that a kind of ABR signal wave crest of the present invention detects, the crest that is applied to the ABR ripple detects and Automatic Logos.No matter ABR is sensation level or the signal of hearing level, and the method can be easy to, very accurately detect all crests in the waveform, can analyze individually each ABR wavelet (such as the V ripple).Monitoring for threshold of audibility assessment and surgical operation provides effective ABR parameter.
Description of drawings
Fig. 1 is that vector calculates sketch map in the algorithm that detects of a kind of ABR signal wave crest of the present invention;
Fig. 2 is the ABR signal graph of removing in the algorithm that detects of a kind of ABR signal wave crest of the present invention behind the noise;
Fig. 3 is ABR peak value sign figure as a result in the algorithm that detects of a kind of ABR signal wave crest of the present invention.
The specific embodiment
The algorithm algorithm that a kind of ABR signal wave crest detects comprises the steps:
One, the denoising of original ABR signal:
Adopt the superposed average technology that the original ABR signal that collects is carried out denoising, namely apply N stimulation, the discretization data were adjusted into the moment of each stimulation appearance align, then average.Be that these adopt ABR signal that superposed average methods draw all based on following hypothesis:
A) under identical stimulation, the waveform of brain evoked response signal is constant, and namely response signal is deterministic signal.
B) background noise has randomness, through repeatedly superposeing and can eliminating.And the pass of it and ABR signal is f (t)=s (t)+n (t), and s in the formula (t) is real ABR signal (not Noise), and n (t) is background noise, and f (t) is measurement data (ABR signal+noise).Under such hypothesis, we have obtained relatively level and smooth ABR signal.
Two, the Automatic Logos of ABR crest:
The BR ripple is 2D signal, establishes the sampling time (discrete time) to be x, sampled value (amplitude) is y, namely sampled point be expressed as ( x, y).
For the signal curve that collects, get window area, be made as A, value is respectively y 1 , y 2.。。, y m , the sampling time is respectively x 1 , x 2 ..., x m , altogether get mIt is individual, mBe odd number, the coordinate of each point be ( x n , y n ), wherein n=1,2... mThen the middle point coordinates in the window area is
Figure 2011101266530100002DEST_PATH_IMAGE012
, the vector between other points of window area and the mid point can be expressed as
Figure 68426DEST_PATH_IMAGE002
, the difference of the amplitude between 2 is
Figure 273142DEST_PATH_IMAGE003
, existing along y direction of principal axis amount of orientation (0 ,-△ y), be designated as
Figure 658993DEST_PATH_IMAGE004
, compute vector then
Figure 327872DEST_PATH_IMAGE005
With
Figure 258919DEST_PATH_IMAGE004
The cosine value of angle is:
Figure 13248DEST_PATH_IMAGE006
, vector between each point and the mid point in the calculation window zone
Figure 459183DEST_PATH_IMAGE005
And vector
Figure 982569DEST_PATH_IMAGE004
The included angle cosine value
Figure 350096DEST_PATH_IMAGE007
, and among the accumulative total window A
Figure 591722DEST_PATH_IMAGE007
Value is calculated SUM=
Figure 53796DEST_PATH_IMAGE008
, vector calculates sketch map as shown in Figure 1.
Therefore establish total N the data point of ABR data, window size is m point, and take step-length as 1 moving window, window size is constant, and the window number that then obtains is N-m+1, is designated as A 1, A 2A N-m+1Accumulate successively in each window Value calculates corresponding SUM 1, SUM 2SUM N-m+1, and record the mid point coordinate figure of each window.By the characteristic of cosine function as can be known,
Figure 970116DEST_PATH_IMAGE009
With
Figure 387453DEST_PATH_IMAGE010
Cosine value is being for just,
Figure 466268DEST_PATH_IMAGE011
Cosine value is worked as SUM for negative nSUM N-1, and SUM nSUM N+1, corresponding window A then nThe interior maximum that occurred, maximum are window A nMid point.So we if with the cosine value of each window and SUM by from big to small ordering, come the larger window of front SUM and just comprised all extreme points.Next step work is exactly to have selected ABR ripple I, II, III, IV, V wave-wave peak just in these points.
At Xinhua Hospital 3 individualities without dysaudia are carried out a large amount of ABR experiments, adopt short sound to stimulate, with stimulus frequency 23Hz, under 40dB, 45dB, 50dB, 55dB, 60dB, 65dB, each stimulation sound intensity of 70dB, 75dB, gather 1024 data points in time at 10ms, obtained the experimental data of priori.The original ABR signal that we at first will collect carries out denoising, then adopts new algorithm that it is carried out the crest Automatic Logos.
Rely on the preclinical clinical experience value of ABR wavelet, roughly determine the position at I, II, III, IV, V wave-wave peak according to the preclinical empirical value excursion of each ripple.Each PL reference value of ABR when only select to stimulate sound intensity 75dB here ( ) be reference frame, the PL reference value of ABR when shown in following table one, stimulating sound intensity 75dB:
Table one
Figure 2011101266530100002DEST_PATH_IMAGE014
Corresponding we collect the ABR data at Xinhua Hospital, corresponding 1024 data points of 10ms.Each wavelet corresponding data point scope incubation period is following table two wavelet corresponding data points incubation period:
Table two
The ripple I The ripple II The ripple III The ripple IV The ripple V
167±14 291±17 400±17 513±107 588±20
178±10 282±16 391±16 512±107 582±21
136±17 259±23 374±26 502±26 571±27
According to top form selection window length, length of window should be greater than the preclinical maximum changing range 54 of ripple, and less than the minimum range 101(of every adjacent two wave crest points here because the IV ripple is too unstable, do not consider the IV ripple).Therefore, be that 81 points are proper by our selection window length of upper table.Because I, III, V ripple absolute latency are more stable, we can at first find the wave crest point of I, III, V ripple, then between I and III crest, look for the II wave crest point, between III and V crest, look for the IV wave crest point, at last I, II, III, IV, V crest are identified.
Find the coordinate figure of all crests of ABR ripple according to the method, and then can calculate I, II, III, IV, accurate incubation period of V ripple and amplitude.
Carry out the date processing experiment with MATLAB7.0, MATLAB7.0 has powerful data-handling capacity.At first the ABR initial data is carried out superposed average and process to remove noise, ABR signal behind the removal noise as shown in Figure 2, ABR signal after the denoising Processing is very level and smooth as seen from Figure 2, below we use the algorithm of proposition that all crests of ABR in the experimental data are identified, result such as Fig. 3, a large amount of experiments show that the crest sign that we realize has very high correctness.

Claims (1)

1. the algorithm that the ABR signal wave crest detects is characterized in that, uses vector operation that the ABR crest is carried out Automatic Logos, comprises following concrete steps:
1) denoising of original ABR signal:
Adopt the superposed average technology that the original ABR signal that collects is carried out denoising, namely apply N stimulation, the discretization data were adjusted into the moment of each stimulation appearance align, then average;
Be that these adopt ABR signal that superposed average methods draw all based on following hypothesis:
A) under identical stimulation, the waveform of brain evoked response signal is constant, and namely response signal is deterministic signal;
B) background noise has randomness, through repeatedly superposeing and can eliminating;
And, the pass of it and ABR signal is f (t)=s (t)+n (t), s in the formula (t) is the real not ABR signal of Noise, n (t) is background noise, f (t) is the measurement data of ABR signal+noise, under such hypothesis, obtained relatively level and smooth ABR signal;
2) the ABR ripple is 2D signal, establishes the sampling time to be x, sample magnitude is y, namely sampled point be expressed as ( x, y), for the signal curve that collects, get window area, be made as A, value is respectively y 1 , y 2 , 。。 , y m ,, the sampling time is respectively x 1 , x 2 ..., x m , altogether get mIt is individual, mBe odd number, the coordinate of each point be ( x n , y n ), wherein n=1,2... m, then the middle point coordinates in the window area is
Figure DEST_PATH_IMAGE002
, the vector between other points of window area and the mid point can be expressed as
Figure DEST_PATH_IMAGE004
, the difference of the amplitude between 2 is
Figure DEST_PATH_IMAGE006
, existing along y direction of principal axis amount of orientation (0 ,-△ y), be designated as
Figure DEST_PATH_IMAGE008
, compute vector then
Figure DEST_PATH_IMAGE010
With
Figure 313168DEST_PATH_IMAGE008
The cosine value of angle is:
Figure DEST_PATH_IMAGE012
, vector between each point and the mid point in the calculation window zone
Figure 417259DEST_PATH_IMAGE010
And vector
Figure 196996DEST_PATH_IMAGE008
The included angle cosine value
Figure DEST_PATH_IMAGE014
, and among the accumulative total window A
Figure 529888DEST_PATH_IMAGE014
Value is calculated SUM=
Figure DEST_PATH_IMAGE016
3) establish total N the data point of ABR data, window size is m point, and take step-length as 1 moving window, window size is constant, and the window number that then obtains is N-m+1, is designated as A 1, A 2A N-m+1Accumulate successively in each window Value calculates corresponding SUM 1, SUM 2SUM N-m+1, and record the mid point coordinate figure of each window; By the characteristic of cosine function as can be known,
Figure DEST_PATH_IMAGE018
With
Figure DEST_PATH_IMAGE020
Cosine value is being for just,
Figure DEST_PATH_IMAGE022
Cosine value is worked as SUM for negative nSUM N-1, and SUM nSUM N+1, corresponding window A then nThe interior maximum that occurred, maximum are window A nMid point;
4) with the cosine value of each window and SUM by from big to small ordering, come the larger window of front SUM and just comprised all extreme points; In these points, select the crest of ABR ripple I, II, III, IV, V.
CN 201110126653 2011-05-17 2011-05-17 Brand-new algorithm for ABR (auditory brainstem response) signal crest detection Expired - Fee Related CN102217932B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN 201110126653 CN102217932B (en) 2011-05-17 2011-05-17 Brand-new algorithm for ABR (auditory brainstem response) signal crest detection

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN 201110126653 CN102217932B (en) 2011-05-17 2011-05-17 Brand-new algorithm for ABR (auditory brainstem response) signal crest detection

Publications (2)

Publication Number Publication Date
CN102217932A CN102217932A (en) 2011-10-19
CN102217932B true CN102217932B (en) 2013-04-03

Family

ID=44774833

Family Applications (1)

Application Number Title Priority Date Filing Date
CN 201110126653 Expired - Fee Related CN102217932B (en) 2011-05-17 2011-05-17 Brand-new algorithm for ABR (auditory brainstem response) signal crest detection

Country Status (1)

Country Link
CN (1) CN102217932B (en)

Families Citing this family (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103565427B (en) * 2013-11-19 2016-09-28 深圳邦健生物医疗设备股份有限公司 Paracycle physiological signal characteristic point detection
CN109431487B (en) * 2018-09-28 2021-11-16 上海乐普云智科技股份有限公司 Method for identifying typical data of ventricular heart beat in electrocardiogram dynamic real-time analysis data
CN109259750B (en) * 2018-11-12 2023-04-11 浙江清华柔性电子技术研究院 Heart rate calculation method and device, computer equipment and storage medium
CN110074782B (en) * 2019-04-30 2020-07-31 上海交通大学医学院附属第九人民医院 Data automatic processing method based on adaptive averaging method in auditory brainstem response test
WO2020249069A1 (en) * 2019-06-13 2020-12-17 上海交通大学医学院附属第九人民医院 Electrophysiological test method for auditory brainstem implant and recording electrode used by method
CN111623958B (en) * 2020-05-18 2021-11-12 长春欧意光电技术有限公司 Wavelet peak-peak value extraction method in interference signal
CN112971776A (en) * 2021-04-19 2021-06-18 中国人民解放军总医院第六医学中心 Method and device for determining position of characteristic waveform in hearing detection waveform

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5697379A (en) * 1995-06-21 1997-12-16 Neely; Stephen T. Method and apparatus for objective and automated analysis of auditory brainstem response to determine hearing capacity
EP1788937A2 (en) * 2004-09-16 2007-05-30 Everest Biomedical Instruments Method for adaptive complex wavelet based filtering of eeg signals
CN101014283A (en) * 2004-05-01 2007-08-08 Bsp生物信号处理有限公司 Apparatus and method for analysis of high frequency qrs complexes
CN101856225A (en) * 2010-06-30 2010-10-13 重庆大学 Method for detecting R wave crest of electrocardiosignal

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US7223246B2 (en) * 2003-06-06 2007-05-29 House Ear Institute Diagnosis of the presence of cochlear hydrops using observed auditory brainstem responses

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5697379A (en) * 1995-06-21 1997-12-16 Neely; Stephen T. Method and apparatus for objective and automated analysis of auditory brainstem response to determine hearing capacity
CN101014283A (en) * 2004-05-01 2007-08-08 Bsp生物信号处理有限公司 Apparatus and method for analysis of high frequency qrs complexes
EP1788937A2 (en) * 2004-09-16 2007-05-30 Everest Biomedical Instruments Method for adaptive complex wavelet based filtering of eeg signals
CN101856225A (en) * 2010-06-30 2010-10-13 重庆大学 Method for detecting R wave crest of electrocardiosignal

Also Published As

Publication number Publication date
CN102217932A (en) 2011-10-19

Similar Documents

Publication Publication Date Title
CN102217932B (en) Brand-new algorithm for ABR (auditory brainstem response) signal crest detection
CN105919584B (en) Heart rate method of estimation and device for wearable heart rate monitor apparatus
Li et al. Detection of ECG characteristic points using wavelet transforms
Gómez-Herrero et al. Automatic removal of ocular artifacts in the EEG without an EOG reference channel
CN109907752B (en) Electrocardiogram diagnosis and monitoring system for removing motion artifact interference and electrocardio characteristic detection
CN105105737B (en) Motion state rhythm of the heart method based on photoplethaysmography and spectrum analysis
CN102440768B (en) Method for extracting pulse waveform feature points
CN104545863B (en) BCG hearts rate extracting method and system based on Fuzzy Pattern Recognition
CN103761424A (en) Electromyography signal noise reducing and aliasing removing method based on second-generation wavelets and ICA (independent component analysis)
CN104305992B (en) A kind of interactive fast automatic extracting method of Fetal ECG
CN106618542A (en) Denoising heart rate detecting device and method
Abbaspour et al. Evaluation of wavelet based methods in removing motion artifact from ECG signal
CN103932687B (en) Method and device for preprocessing pulse condition signal
CN112957056B (en) Method and system for extracting muscle fatigue grade features by utilizing cooperative network
CN103829944B (en) Based on the thoracic impedance signal processing method of pattern recognition
CN107361764A (en) A kind of rapid extracting method of electrocardiosignal signature waveform R ripples
Belkadi et al. A robust QRS detection approach using stationary wavelet transform
CN110680307A (en) Dynamic blood pressure monitoring method based on pulse wave conduction time in exercise environment
CN110236538B (en) Automatic real-time detection method for motion starting point based on electromyographic signal
Kaur et al. An efficient R-peak detection using Riesz fractional-order digital differentiator
Elbuni et al. ECG parameter extraction algorithm using (DWTAE) algorithm
Pan et al. Detection of ECG characteristic points using biorthogonal spline wavelet
Nair et al. Adaptive wavelet based identification and extraction of PQRST combination in randomly stretching ECG sequence
Heydari et al. Adaptive wavelet technique for EEG de-noising
CN104778342A (en) Heart sound characteristic extraction method based on wavelet singularity entropy

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
C14 Grant of patent or utility model
GR01 Patent grant
CF01 Termination of patent right due to non-payment of annual fee

Granted publication date: 20130403

Termination date: 20160517

CF01 Termination of patent right due to non-payment of annual fee