The proposed work addresses the growing interest in fall detection among researchers, which is driven by its wide range of applications in healthcare and safety monitoring. This research uses wearable sensors like accelerometers and gyroscopes to evaluate human behavior, with a particular emphasis on fall detection. Effective fall detection requires the extraction of crucial temporal elements from gathered data. Traditional methods are frequently computational-expensive, necessitating specialized knowledge and including extensive feature engineering and data processing. This paper introduce Ensemble learning Model using Convolutional Neural Network (CNN) and Bidirectional Long Short-Term Memory (BiLSTM) of deep learning model, marking a significant development. This novel method of selection of spatial and temporal separately fusion of both spatial and temporal which make and accurately with of 98.17% on the Federated Research Data Repository (FRDR). By automating feature detection and streamlining the data processing pipeline, this approach improves the efficiency and effectiveness of fall detection systems, paving the way for more robust and scalable solutions in real-world applications.