Optimized fall detection using hybrid BiLSTM BiGRU additive attention model and BAOA driven feature selection system

With the increasing aging population and the prevalence of fall-related injuries among the elderly, the need for precise and reliable fall detection systems has become crucial. Deep learning techniques have been used to play a transformative role in fall detection domain due to their ability to learn complex patterns and relationships in data. However, these methods also face significant challenges and limitations. In this paper we have proposed a novel fall detection method utilizing a hybrid model combining Bidirectional Long Short-Term Memory (BiLSTM) and Bidirectional Gated Recurrent Unit (BiGRU) networks, enhanced with additive attention mechanism and an optimized feature selection process utilizing the Binary Arithmetic Optimization Algorithm (BAOA). The proposed hybrid BiLSTM-BiGRU-Additive Attention Model and BAOA Driven Feature Selection system is designed to proficiently capture temporal dependencies in the data. The model is evaluated on three datasets collected using wearable sensors, namely SisFall, UMAFall, and UP-Fall, experimental results demonstrate that the proposed model significantly outperforms traditional deep learning architectures, and non-optimized versions, achieving accuracies of 99.50%, 99.85%, and 99.68% for the SisFall, UMAFall, and UP-Fall datasets, respectively making our proposed model suitable for real-time fall detection systems.