Based On Wavelet Decomposition And Teager Energy Operator P300 Feature Extraction And Classification Algorithms

EEG is one of the most common and important human bioelectricity, according to research shows that its contain rich physiological, psychological and pathological information. Brain machine interface(brain-computer interface,BCI) is a kind of communication interface in the field of life sciences, studies have shown that by record EEG and through certain treatmented can interpret human brain thinking, thus able to convert them into the machines can identify control command so as to realize the human brain control computer, wheelchair, household appliances and robot, so researchers to study brain electric signal has the vital significance, mainly used in brain function, clinical neurologist disease analysis, judge people thinking and defense industries, etc.Based on visual evoked potentials P300 brain-machine interface is easy to implement, so this paper is to study how to determine obtained visually evoked EEG data were contain evoked potentials P300 and its processing method:].In EEG acquisition system, the power frequency interference is one of the most common interference, the method mandatory remove power frequency interference through hardware will remove useful EEG, but if uses an algorithm to soft remove power frequency interference can reduce loss of EEG information. Therefore, when preprocessing the EEG signal, this paper specially designed a power frequency notch filter, filter out 50Hz frequency interference, it can ensure other frequencies signal no losing, and effectively suppress power frequency interference in the input signal;2.the EEG data which after Power frequency filter may still contain other noise, and even submerged by noise, so the extracted EEG data features brought great difficulties, so to remove the noise and to prepare for the feature extraction has great significance. Because of the wavelet analysis is a kind of window size constant, variable size of typical in both time domain and frequency domain has good localization signal analysis method, it can adaptively according to the characteristics of being analyzed signal analysis of different scales are analyzed signal, so this paper use based on"sym6"wavelet analysis to the default threshold denoising.3.In order to improve the speed of resolution and extracting feature amount of EEG data time domain characteristics, the common practice is preserve the basic characteristics of EEG data when the EEG dimensionality reduction. According to the introduction and analysis of the wavelet decomposition, this paper adopted based on"sym6"wavelet basis decomposition EEG data, a EEG data sampling point 206 after wavelet decomposition dimension reduction to 23 points.4.In order to obtain more P300 characteristics have reached training better support vector machine (SVM) classification model, this paper puts forward the EEG data energy domain characteristics. Thus introduces Teager energy operator, Teager energy operator computation is a simple calcuation, and calculate the EEG data more quickly.5. After extract EEG data energy domain and time domain features, to use a Binary classification methodology to determine whether it contains P300. this paper proposes the use of two-class support vector machine, the choice based on SVM RBF core functions EEG data model can analyzing whether it contains P300 meet the requirements.Therefore, in this paper, based on wavelet decomposition and Teager energy operator comprehensive feature extraction algorithm accuracy is higher than the algorithm based on time domain feature extraction. Has a good theoretical research and application value.