Time-Frequency Feature Enhancement of Radar Signal Using 2D Gaussian Filter
Khairul Khaizi Mohd Shariff, Suraya Zainuddin, Aziati Husna Awang, Mohd Khairil Adzhar Mahmood · 2021
Millimetre-wave radar plays an increasing role in wireless sensing for hand gesture recognition. Despite its huge potential, several difficulties still exist, including the problem of the poor and unreliable feature of radar micro-Doppler signal due to the weak hand reflectivity and noisy environment. Therefore, enhancing the features will result in higher recognition accuracy. In this paper, we proposed a simple scheme to enhance the micro-Doppler feature. Specifically, the radar micro-Doppler signal is transformed to a time-frequency domain, and we reduce the effect of noise by simple spectral subtraction and squared magnitude. Finally, we apply a Gaussian filter to smooth spectra magnitudes, thereby enhancing the signature pattern from the background noise. Experimental results using a real hand gesture dataset show that the proposed scheme achieved a high classification accuracy of 97.5%.