A T-CNN time series classification method based on Gram matrix

Abstract

Time series classification is a basic task in the field of streaming data event analysis and data mining. The existing time series classification methods have the problems of low classification accuracy and low efficiency. To solve these problems, this paper proposes a T-CNN time series classification method based on a Gram matrix. Specifically, we perform wavelet threshold denoising on time series to filter normal curve noise, and propose a lossless transformation method based on the Gram matrix, which converts the time series to the time domain image and retains all the information of events. Then, we propose an improved CNN time series classification method, which introduces the Toeplitz convolution kernel matrix into convolution layer calculation. Finally, we introduce a Triplet network to calculate the similarity between similar events and different classes of events, and optimize the squared loss function of CNN. The proposed T-CNN model can accelerate the convergence rate of gradient descent and improve classification accuracy. Experimental results show that, compared with the existing methods, our T-CNN time series classification method has great advantages in efficiency and accuracy.

Device Identity Recognition Based on an Adaptive Environment for Intrinsic Security Fingerprints

“…To better handle these extracted features, we introduce Gramian angular fields (GAFs) [12] for feature transformation. The use of Gramian angular fields enables the conversion of time series data into an image-like format, effectively capturing complex relationships between features and providing richer input information for subsequent hypergraph neural network models.…”

“…G GASF and G GADF represent the relative relationships between different vectors in terms of superimposition and difference over time intervals, reflecting their temporal dependencies [12].…”

Identification of runner fatigue stages based on inertial sensors and deep learning

“…CNN can extract effective features from signals and has achieved good results in speech recognition, image classification, and text analysis. It has been shown that CNN can retain the correlation between the before and after signals in human action recognition relative to other models when classifying time-series data in human action recognition (Wang et al, 2022). LSTM is a typical representative of recurrent neural networks, which has been widely used in fields such as handwriting recognition (Graves et al, 2009), character generation (Ergen and Kozat, 2017), automatic language translation (Tang et al, 2020), speech recognition (Coto-Jiménez, 2019), video processing (Song et al, 2019), and so on.…”