The Extension Study Of ECG Identification To Its Motion State

Because of its universality, uniqueness, stability and measurability, ECG is being widely used for identification.But the studies of ECG identification before only focus on the feature extraction and pattern recognition under the rest state. However, in practical application, it often comes to the situation that one person need to be identified when he is under movement or just finished it. Therefore, we must consider the expansion of identification in the motion state, namely it is required to identify the individual accurately both in rest and motion state. But the exercise would have influence on the measurement of signal, making it more difficult to locate the feature points. It also leads to the changes of feature vectors used before, which affects the effect of identification.This thesis mainly improves two aspects of the original ECG identification technology to make it suitable for excise occasions. First, a new method of R-wave detection using the threshold of slope between maxima has been proposed, which is insensitive to the changes in amplitude of peak. Second, using fusion features of morphological characteristics and KPCA makes two types of feature vectors play a complementary role.The thesis firstly verifies the validity of the new method of R-wave detection through experiments. Then extract the feature vectors. Finally, a comparative experiment proves that fusion features has better performance than morphological characteristics and KPCA under different motion state, which verifies the effectiveness of the improvement.