A Lightweight Neuromorphic CNN for Human Activity Recognition on Edge Device

Human activity recognition (HAR) is a rapidly growing field with numerous practical applications, such as healthcare, assistive living, and security. The majority of HAR applications are ideally suited for edge deployment. However, edge devices have limited memory and energy, making it challenging to deploy HAR tasks on them. To address this issue, a biological brain optimized neuromorphic Convolution Neural Network (biCNN-HAR) for HAR is developed. The performance of biCNN-HAR is evaluated on mHealth dataset, and its results are compared with baseline CNN models and state-of-the-art models. The findings demonstrate that biCNNHAR outperforms other models while utilizing fewer computational resources.