Human Activity Recognition Algorithm Based on One-Dimensional Convolutional Neural Network

Abstract

Human activity recognition (HAR) is widely used in healthcare, personal fitness, physical training and military, etc. How to distinguish various human activities accurately (such as running, walking, walking upstairs and downstairs, jumping and standing) has become an important problem in human-computer interaction. The computer vision method requires a large amount of computing resources, and it is not highly accuracy and can be easily disturbed by other objects in the background. The sensor-based method can achieve high accuracy, and it requires few computing resources, and is not disturbed by the background. This paper proposes a method based on the one-dimensional convolutional neural network (1D-CNN) to classify the sensor signals of some different activities. For comparison, this paper applies some widely used methods to accomplish the recognition task with the same dataset. Then, it tests the proposed 1D-CNN model with different datasets, for the purpose of testing its generality across users. The experimental results show that the proposed model achieves an accuracy of 95.12% with the said datasets, which is higher than those of the other methods by about 8% on average. This indicates that the proposed method has good performance in terms of generality across users, and at the same time provides a higher accuracy. The obtained results can improve the accuracy of current technologies.

AksharaNet: A GPU Accelerated Modified Depth-Wise Separable Convolution for Kannada Text Classification

“…DSC used with SqueezeNet as encoder and Depth-wise Separable transpose convolution as a decoder resulted in much lesser parameters. Lately 1D-CNN [19] used single dimension convolution to classify sensor signals as DiCENet [20] unit, that is built using dimension-wise convolutions and dimension-wise fusion have proved as an efficient performer.…”