Remote Speech Decryption Using Millimeter-Wave Micro-Doppler Radar

This work proposes a novel non-contact methodology for human speech recognition using micro-Doppler radar and advanced signal processing. A 94 GHz millimeter-wave radar tracks micro-vibrations of vocal cords during speech. The radar echoes undergo a short-Time Fourier transform (STFT) to generate spectrographic representations capturing time-varying micro-Doppler signatures. Optimal STFT window lengths balance frequency resolution and temporal precision for accurate visualization of vocal micro-motions. The concept is first validated by modeling vocal cord vibrations using a piezoelectric crystal driven by single-Tone, frequency-modulated, and audio signals. Experimental results demonstrate reliable tracking of abrupt 2 kHz frequency changes, indicating the potential for discerning human vocalization patterns through obstructions. Overcoming limitations of acoustic/vision-based approaches, this privacy-preserving methodology enables secure remote speech processing for defense, security, and surveillance applications by analyzing only abstract vocal signatures.