Enhanced Real-Time Surveillance and Suspect Identification Using CNN-LSTM Based Body Language Analysis

The exponential rise in criminal activities necessitates advanced methods for suspect identification and surveillance. This research aims to tackle this issue through the development of a sophisticated video analytics system leveraging computer vision and deep learning. The primary objective is to accurately identify suspects based on body language patterns extracted from video inputs. We propose a CNN-LSTM based Body Language Rule System (BLRS) that integrates Convolutional Neural Networks (CNNs) for spatial feature extraction and Long Short-Term Memory (LSTM) networks for temporal sequence learning. The system processes video frames to identify key body language indicators such as gestures, postures, and facial expressions. Extensive evaluations using the UCF-Crime Dataset demonstrate the model's high accuracy, with a precision of 95.5%, recall of 95.7%, and overall accuracy of 95.3%. The results indicate that the BLRS significantly outperforms traditional human action recognition models, providing robust and reliable identification of suspicious activities. This research concludes that integrating CNN and LSTM networks within a unified framework enhances real-time surveillance capabilities. The proposed system holds substantial potential for improving public safety and security by enabling more effective monitoring and identification of suspects through advanced body language analysis.