The Identification Robustness Research Of ECG

Today Internet of things is popularized rapidly, ECG signal identification applying in wearable devices has great prospects. Due to the characteristics of mobile devices, ECG signal acquisition device should have high integration; data is cruder than data achieved by accurate medical equipment. And when we collect ECG signal by the wearable devices, the finger or wrist need to touch the acquisition terminal, the contact point may be loose. As collectors, they cannot collect data in the peace mode every time, when they collect ECG data after they do some sport, the collecting data will be different than the data collecting in the peace mood. These entire situations will cause the identification fault with the poor robustness of identification algorithm. The noise of ECG data will be big, jitter data will show up and the heart rate will change greatly, and so on. For that the robustness of ECG authentication algorithm has higher requirements. In order to solve these problems this paper, we will start with two aspects that is solving the problem of movement and increase the dimension of ECG signal to enhance the robustness of ECG identification algorithm, and multistage ECG authentication algorithm is putting forward.This paper firstly puts forward multi-stage identification algorithm based on motion problems. In order to solve the problem of heart rate variability of collecting data, adding the processing of storing polymorphic average template. Then extract the all-round feature of ECG signal as input of multi-layer authentication algorithm. In the training stage, we should train the feature to obtain best threshold with different threshold selection strategy. Then input test set into multi-stage authentication algorithm to identify status according to the best threshold. Each stage of the input samples are authentication error samples in last stage, it can reduce the number of sample of the each layer, and it can identify the samples which are difficult to identify in last stage by the feature of other angle in the next stage. The experimental results show that multi-stage authentication algorithm improves the accuracy compared with single authentication. And compared with single feature, the multi-stage authentication algorithm with a variety of characteristics has more significant effect.To further enhance the robustness of ECG identification this paper put forward multi-stage identification algorithm based on multi-dimensional ECG signal. At first fusing the multi-dimensional ECG signal in the data layer. When double lead ECG signal is mapped into a two dimensional space, we will get a sparse matrix, take the sparse matrix as a binary image, then extract a variety of feature, such as overall appearance, wavelet coefficients, shape feature and density distribution characteristics. Train the characteristics by different threshold selection strategy, then input the testing set and the best threshold in multi-stage identification algorithm for authentication. In the process of experiment, the unidimensional data in single stage identification algorithm and multidimensional data in single stage identification algorithm are compared. The results show that the proposed multi-dimensional ECG signal of multi-stage authentication algorithm’s accuracy can top up 99.72 %. At last we propose multi-template multi-dimensional multi-stage identification algorithm to solve the sport problem of multi-dimensional ECG signal. The accuracy of algorithm which applies on the database collected by ourselves is up to 93.75%. At the end of this paper we evaluate the all algorithms by the time complexity.