Human Activity Recognition (HAR) such as fall detection has become increasingly critical due to the aging population, necessitating effective monitoring systems to prevent serious injuries and fatalities associated with falls. This study focuses on fine-tuning the Vision Transformer (ViT) model specifically for HAR using radar-based Time-Doppler signatures. Unlike traditional image datasets, these signals present unique challenges due to their non-visual nature and the high degree of similarity among various activities. Directly fine-tuning the ViT with all parameters proves suboptimal for this application. To address this challenge, we propose a novel approach that employs Low-Rank Adaptation (LoRA) fine-tuning in the weight space to facilitate knowledge transfer from pre-trained ViT models. Additionally, to extract fine-grained features, we enhance feature representation through the integration of a serial-parallel adapter in the feature space. Our innovative joint fine-tuning method, tailored for radar-based Time-Doppler signatures, significantly improves HAR accuracy, surpassing existing state-of-the-art methodologies in this domain. Our code is released at https://github.com/wangyijunlyy/SelaFD.

Millimeter-wave radar has attracted increasing attention for respiratory monitoring due to its non-contact operation and privacy-preserving characteristics. Nevertheless, extracting fine-grained respiratory waveforms from non-stationary radar signals remains highly challenging, as these signals are frequently contaminated by various interferences, most notably aperiodic body micromotion. The spectral components of such interference often overlap with the respiratory frequency band and typically exhibit power levels that significantly exceed the target signal. This letter introduces the Gated Attention Wavelet Network (GAWNet), an interpretable framework that integrates deep learning with physical priors by operating on radar phase information in the wavelet domain. GAWNet leverages a two-stage suppression strategy: first, a Temporal Gated Attention (TGA) encoder combines convolutional gating and self-attention to achieve initial interference reduction; second, a Frequency Gated Attention (FGA) decoder provides further refinement by transforming wavelet coefficients to the frequency domain for precise filtering. The clean respiratory waveform is then reconstructed using an Inverse Discrete Wavelet Transform (IDWT). Extensive experiments with data from 12 subjects demonstrate that GAWNet consistently outperforms state-of-the-art models and exhibits robust generalization capability.