Deep Learning-Based Human Activity Recognition Algorithms: A Comparative Study

Human activity recognition is essential for developing personalized healthcare, fitness tracking, and fall detection systems. This study focuses on developing an accurate deep learning-based algorithm for recognizing human activities using accelerometer and gyroscope data. The proposed model is based on a 1D-CNN-BiLSTM architecture and is trained on a public dataset. In this study, six different human actions-walking, walking upstairs, walking downstairs, sitting, standing, and laying-are identified to examine how well the proposed model performs compared to a baseline model and to assess how well the model recognizes these activities. According to the results, the suggested model performs better than the baseline model at identifying human activity and is more accurate overall. All six activities can be recognized by the suggested model with high accuracy, with walking, walking downstairs, and walking upwards having the best performance. The model's success is attributed to its ability to capture both spatial and temporal dependencies in the data, allowing it to effectively recognize human activities based on accelerometer and gyroscope data. The study's findings have significant implications for developing activity recognition systems, especially for real-time health monitoring, fall detection, and fitness tracking applications. The proposed model's high accuracy can improve the performance of such systems, leading to better outcomes for patients and individuals. The study's limitations are also discussed, including the relatively small size of the dataset and the lack of variability in the activities performed. This study provides evidence of the effectiveness of a deep learning-based algorithm for recognizing human activities using accelerometer and gyroscope data. It demonstrates the potential of such models for developing accurate and reliable activity recognition systems with important applications in healthcare and fitness tracking.