A Novel CNN-SVM Hybrid Model for Human Activity Recognition

Human activity recognition (HAR) is a growing field that focuses on using sensor data to automatically detect and analyze human motions. HAR has gained a lot of attention and interest due to its extensive use in many different domains, such as healthcare, sports, security, and many others. In this paper, we propose a hybrid approach for HAR that combines feature extraction using a Convolutional Neural Network (CNN) and classification using a Support Vector Machine (SVM). The preprocessing step is performed using Linear Discriminant Analysis (LDA). LDA transforms the input data from its original high-dimensional space to a new, lower-dimensional space, resulting in enhanced separation between different classes. By passing the transformed features to the CNN, the network can extract more relevant and discriminative features compared to directly using the raw sensor data. This leads to improved classification and prediction performance. The proposed approach is evaluated on two widely used datasets, achieving an accuracy of 96.30% on the UCI HAR and 95.00% on PAMAP2.