Figure 2. Performance characterization of REAL. a) REAL collects audios from the speaker. b) STFT spectrograms of ‘Ah~~’ recorded by microphone, REAL on throat and REAL on mask. c) Frequency cross-section of the signals in b) at time 1s. d) SNR characterization of REAL system from a mask on a loudspeaker playing 500 Hz sine audio. e) The relationship between laser power and total noises, which includes electronic noise ( \(σ_e\) ), shot noise ( \(σ_s\) ) and laser power fluctuation noise ( \(σ_l\) ). f) The REAL SNR as a function of measurement distance, with different laser power ( \(P\) , left) and vibration magnitude ( \(∆r\), right)