juseljustj/human-activity-recognition-using-deep-learning-techniques

Human-Activity-Recognition-using-Deep-Learning-Techniques

Human activity recognition (HAR) is an emerging area of research in deep learning that focuses on predicting human activity based on sensor data. The goal of HAR is to recognise patterns of behaviour associated with sitting, standing, walking, and falling in elderly individuals using state-of-the-art deep learning techniques, including Convolutional Neural Networks (CNNs), Long Short-Term Memory (LSTMs) and a hybrid CNN-LSTM. The classification of human activities will be based on infrared imaging data collected from low-resolution Grid-Eye IR Sensors, resulting in a 4-class problem. These sensors can detect the heat emitted by the human body, making them an ideal technology for monitoring activity in a non- intrusive way. The importance of HAR systems in the elderly population cannot be overstated, as the World Health Organization estimates that there will be 1.5 billion people aged 65 years or over worldwide by 2050(World Health Organization, 2022). HAR systems use sensors to track the movements and activities of individuals within their homes. By analysing this data, the system can recognise patterns of behaviour and detect any deviations from the norm that may indicate a problem. This information can then be used to alert caregivers or emergency services, helping to ensure the safety and well-being of elderly individuals who wish to live independently in their homes. Continuous HAR is a non-intrusive technology that can be used to detect human activity without being intrusive or violating privacy. This is especially important for elderly populations who may be reluctant to use intrusive technologies such as surveillance cameras and wearable sensors. In this project, we aim to develop a non-intrusive and continuous HAR system for the elderly population using Grid-Eye IR Sensors. Human activity recognition using IR sensors is a time series problem that can be solved with various neural network architectures. This project will use CNNs, LSTMs, and a hybrid approach that combines both architectures known as CNN-LSTM. These architectures can extract both local and global features from the data, allowing the model to learn patterns of human activity over time and achieve successful classification and anomaly detection.(Fan et al., 2017) Overall, this project demonstrates the potential of deep learning techniques in developing non-intrusive and continuous HAR systems for the elderly population. The system developed in this project has the potential to improve the quality of life for elderly individuals by allowing them to live independently in their homes while ensuring their safety and well-being.