Image Classification of Time Series Based on Deep Convolutional Neural Network

Convolutional Neural Networks (CNN) have achieved great success in image recognition tasks by automatically learning hierarchical feature representations from raw data. Most time series classification (TSC) mainly focuses on one-dimensional signals. In this paper, we plan to study high dimensional time series data. The main idea is following: Firstly, data enhancement is done which means that we use synthetic minority oversampling technique (SMOTE) to preprocess the arrhythmia data. Secondly, we use recurrence plot (RP) to convert the time series into two dimensional texture images. Thirdly, the deep CNN classifier is used for recognition. And finally, time series classification can be regarded as the texture image recognition task. The arrhythmia data is used to demonstrate the effectiveness of the proposed method for processing time series data sets.