AmbiEar

Millimeter wave (mmWave) based sensing is a significant technique that enables innovative smart applications, e.g., voice recognition. The existing works in this area require direct of the human's near-throat region and consequently have limited applicability non-line-of-sight (NLoS) scenarios. This paper proposes AmbiEar, first mmWave recognition approach applicable NLoS AmbiEar on insight causes correlated vibrations surrounding objects, regardless position posture. Therefore, regards objects as ears can perceive sound realizes indirect by vibration objects. By incorporating designs like common component extraction, signal superimposition, encoder-decoder network, tackles challenges induced low-SNR distorted signals. We implement commercial radar evaluate its performance under different settings. experimental results show has word accuracy 87.21% scenarios reduces error 35.1%, compared to approach.