Gang Zheng

ECG Based Identification by Deep Learning

Strategies were proposed for Electrocardiogram (ECG) based identification. Firstly, a selecting mechanism based on information entropy was used to obtain whole heart beat signal; Secondly, a Depth Neural Network (DNN) based on Denoising AutoEncoder (DAE) was adopted in feature selection unsupervised, by which, the robustness of the recognition system could be improved in recognizing. Finally, 98.10% and 95.67% recognition rate were obtained on self-collected calm and high pressure data sets respectively, and 94.39% rate on combined data sets of MIT arrhythmia database (mitdb) and self-collected data averagely.

The Study of Feature Selection Strategy in Electrocardiogram Identification

Identification based on electrocardiogram (ECG) is an emerging hot spot in biometric identification. Feature selection is one of the key research points on it. In the paper, features are firstly calculated from fiducial points of ECG. Secondly, the initial feature set is composed of amplitude, interval, slope, area and some clinical indexes. Thirdly, a feature selection strategy is proposed. The strategy uses stepwise discriminant analysis to calculate the contribution (weight) of each feature for ECG identification. On the basis of contribution sorting, accumulative recognition rate is calculated. Furthermore, a key feature subset for ECG identification is acquired when accumulative recognition rate reaches a steady level. Fourthly, the identification procedure works on key feature subset. ECG data from both PTB and laboratory is used in experiments. Experimental results show that the identification accuracy of the two data sets is 99.7% and 94.8% respectively.

A Modified K-NN Algorithm for Holter Waveform Classification Based on Kernel Function

Several problems are existed when K-NN (K- nearest neighbor) method is used to classify the Holter waveforms: the data scale is too large; the classification algorithm needs training samples; the K-NN is a linear classification method. Therefore, this paper proposes a new K-NN algorithm; the algorithm is based on kernel function. Through this change, classification is transformed from linear to non-linear. The max-min distance algorithm and k-means clustering algorithm are used to form the training sample set for the modified K-NN algorithm. By this method, Holter waveforms are classified more correctly and automatically.

ECG Based Biometric by Superposition Matrix in Unrestricted Status.

The paper proposed an Electrocardiogram (ECG) feature extraction method for biometric. It relied on ECG superposition number matrix built by several single heartbeat ECG data. The target of the study was to find stable features of the ECG signal under unrestricted status for biometric. By matrix segmentation and similarity comparison, the stable feature distribution was gotten, and stable feature sets were also constructed. 13 volunteers’ ECG data collected by self-made ECG device in different status were gotten, the collecting period was lasting for half year; 28 healthy individuals’ ECG data under calm status were also collected; Besides that, 14 subjects’ ECG data in MIT-BIH were also involved in study. From the result of experiments, the average True Positive Rate (TPR) reached 83.21%, 83.93% and 80% on MIT data set, ECG data set in calm status and ECG data in different status respectively. It is also found that along with the increasing amount of ECG single heartbeat used to build superposition matrix, the stable features of one’s ECG were gradually revealed and this helped ECG based biometric effectively.

A Method of ECG Identification Based on Weighted Correlation Coefficient

When the correlation coefficient (CC) method is used in electrocardiograph (ECG) identification, the accuracy of identification can be affected by the number of templates and different representative templates. In this paper, a template selection method is proposed based on contribution rate of each ECG waveform in the data set, and the second-order differential threshold value is used to determine the number of templates and select representative ECG templates for each individual. The weighted correlation coefficient method is proposed for ECG identification, and weights are calculated by the sorted sequence of contribution rate. The performance of the presented method is tested on the MIT-BIH ECG data set and the hand ECG data collected in real scenery. Comparing with the traditional correlation coefficient method, experimental results show that the average identification accuracies are increased 10.52% and 3.85% respectively when using the presented method.

Study of Key Techniques on Electrocardiogram Printouts Digitizing

Key techniques were proposed in printout electrocardiogram (ECG) digitalization, which were composed by image edge detecting, angle adjusting, and ECG waveform extracting, de-noising, scaling, and saving. The requirements of digitalizing procedure contained ECG amplitude accuracy rate, heart rate counting, and morph changes. Experiments showed that the key techniques can preserve the fatal features and parameters efficiently, precisely, and automatically. The accuracy rate of waveform amplitude reaches 95%, and heart rate of that reaches 98%. The results satisfied the clinical requirements, and can be used for network medical service.

Study of Tunnel Morphin the Similarity Analysis of Time Series Bio-signal Data

One similarity measurement strategy for bio-signal waveform is presented in this paper. A tunnel morph is introduced to describe the waveform. It measures the waveform not only by a value, but considers the curve feature of waveforms. Then a set of integrity definitions are presented in the paper. It contains tunnel construction, tunnel morph width, edge detection, class judgment. A case study on AECG (Ambulatory Electrocardiogram) is also presented to give a proof for the similarity measurement strategy. Through experiments, each type of dataset could be integrated into a tunnel, then, every unknown data whether belongs to this tunnel can be evaluated. Data used in the paper is from MIT/BIH.

A Detection Method for Overlapped Spikes

A new spike detection method is proposed in order to detect the overlapped spikes. In order to avoid missing overlapped spikes, the method adds threshold detection based on window detection method. Moreover, nonlinear energy operator is introduced to make the method strong even under low signal-to-noise ratio situation. In addition, the method solves the repeated detection problem by estimating slopes. Experiments show that the method is good for any occasion whatever the low signal-to-noise ratio or baseline wander. Especially for the overlapped spikes detection, it has much lower false-negative-rate than other traditional detection methods.