Human Activity Recognition using Teknomo-Fernandez Kernelized Discriminant

Abstract

Human activity recognition (HAR) is a crucial problem in the field of human-computer interaction, with applications in various domains such as healthcare, surveillance, and robotics. Traditional machine learning approaches for HAR often rely on hand-crafted features and manual tuning of hyper parameters, which can be time-consuming and limit the accuracy of the recognition system. Recently, deep learning techniques have shown promising results in HAR, but they often require large amounts of labeled data and can be computationally expensive. This paper proposes a novel approach to HAR using Teknomo-Fernandez kernelized discriminant analysis (KF-D) based connectionist deep multilayer perceptron (CDMLP) neural learning. The proposed approach combines the strengths of kernel methods and deep learning to learn robust and efficient representations of human activities. The KF-D method is used to extract features from raw sensor data, which are then fed into a CDMLP network to learn a mapping between the extracted features and the corresponding human activities. The CDMLP network is trained using a back propagation algorithm with a modified version of the cross-entropy loss function. Experiments were conducted on four publicly available datasets, including the Oxford-Hertfordshire Activities of Daily Living (ADL) dataset, the Opportunity dataset, the Human Activity Recognition Using Smart Devices (HARD) dataset, and the WISDM AR Sensor Mining dataset. The proposed approach achieved state-of-the-art performance on all four datasets, outperforming existing methods in terms of accuracy and robustness. The results demonstrate the effectiveness of the proposed approach in recognizing human activities with high accuracy, even in noisy and challenging environments. The proposed approach has potential applications in various domains, including healthcare, surveillance, and robotics. Future work includes extending the approach to recognize more complex human activities and integrating it with other sensors and devices to create a more comprehensive HAR system.