Enhancing Human Activity Recognition with Iterative Adaptive Approach in Assisted Living

Micro-Doppler signatures are a widely used technique for human activity recognition (HAR) in assisted living using radar systems. However, traditional methods such as the Short-Time Fourier Transform (STFT) encounter significant issues, including a compromise between temporal and spectral resolution, susceptibility to noise, and the necessity for precise parameter adjustments. To address these problems, we propose a novel method employing the Iterative Adaptive Approach (IAA) to extract high-resolution micro-Doppler signatures. The proposed method improves temporal resolution by achieving super-resolution in Doppler estimation with fewer samples in the spectral domain. This results in spectrograms with enhanced resolution, thereby improving the accuracy of micro-Doppler classification techniques. The effectiveness of the proposed method is evaluated using a challenging dataset of real-world human activities, which includes six distinct types of activities. Our findings reveal that the proposed IAA-based approach achieves up to a 3.4% improvement in classification accuracy compared to traditional STFT-based methods. This improvement is significant for the research community focused on radar-based HAR and for indoor human activity monitoring in the context of assisted living, offering a more robust and accurate technique for identifying human activities.