A Comparative Study of Classical Machine Learning and Deep Learning Approaches for Human Behaviour Detection Using Multisensor Data

The study investigates the use of machine learning models for human behavior detection using multisensor data, which includes accelerometer, gyroscope, heart rate, temperature, proximity sensors, and other physical data sources. The project aims to create a machine learning model capable of detecting human gestures and behaviors by leveraging these diverse sensor inputs. The dataset includes key features such as movement data, emotional state (mood), activity type, and specific physical gestures like scratching or texting on a phone. The proposed system integrates classical machine learning algorithms, including Random Forest, XGBoost, ADA BOOST,  with a novel deep learning approach for model optimization. The goal is to provide a robust and scalable system for real-time behavior detection, which can be applied to fields like healthcare, human-computer interaction, and surveillance. The system's performance will be evaluated based on prediction accuracy and model interpretability, ensuring its practical deployment in various scenarios. By using advanced sensor data and machine learning, this research aims to provide a more nuanced understanding of human behavior and enhance interaction technologies.

Keywords: Human Behavior Detection , Multisensor Data , Machine Learning ,     , Gesture Recognition , Emotion Detection , Activity Recognition, Real-Time Systems , Gesture Label Prediction , Deep Learning