The Research On Acoustic-Based Passive Sensing Methods On Smart Devices

As a new cutting-edge technology in the past two decades,wireless sensing has created infinite possibilities for many fields such as new human-computer interaction,virtual reality,smart home,health management,medical monitoring,and autonomous driving.In addition to their inherent information-transmitting functions,a wide variety of wireless signals are exploited for sensing to their greatest advantages.The basic principle of wireless sensing technology is that the movement of the target will affect the propagation of the signal in the air.By analyzing the changes of the reflected signal,information such as the distance and speed of the target can be extracted.Acoustic sensing stands out as one of the most promising wireless sensing technologies due to its wide availability and fine sensing granularity.It can extend the simple audio playback and voice control functions of speakers and microphones to higher-level human-computer interaction and health monitoring.It plays a vital role in breaking the chains of COVID-19 transmission and the protecting human lives.Although the passive sensing technology based on acoustic signals has made significant progress in research at home and abroad,the existing technology still faces many challenges in practical applications embodied in the following four aspects: First,when multiple targets are sensed at the same time,the signals reflected from multiple targets are superpositioned and mixed to reach the receiving end.It is difficult to separate these signals to obtain the information of each single target.In the case of a limited number of microphones in commercial smart devices,it is more challenging to separate the signals of multiple targets.Second,weak signal variation and unrobust detection are the key challenges in the sensing of subtle human activities.Under the environment interference and inevitable self-interference,the signal variations caused by subtle movements become more difficult to extract.Third,when sensing a long-range target,the acoustic signal at a high frequency is severely attenuated in the air,so the signals reflected from the target are very weak,making it hard to achieve high sensing performance.Last but not least,when we further expand the sensing range to multiple rooms in indoor environments,the sensing device with a fixed location cannot achieve sensing across physical boundaries.At this time,it is necessary to use a mobile platform to carry sensing equipment.However,the large device movement can easily overwhelm the motion information of the target,which leads to the failure of traditional sensing methods under the device movement scenario.This dissertation integrates model construction and signal processing technology to solve the problems of multi-target interference,limited sensing granularity,limited range,and low sensing accuracy under the device movement scenario in acoustic sensing.The contributions of this dissertation are as follows:(1)To address the problems of separating signals reflected from multiple targets and the large tracking errors in multi-target sensing,we propose a joint estimation algorithm of multidimensional information fusion through theoretical analysis and model verification to reconstruct the reflected signal of each target.We calibrate the tracking errors by combining time domain information,which improves the multi-target resolvability and sensing accuracy.(2)To address the weak signal variation and severe interference problems in the sensing of subtle human activities(e.g.,eye blink detection)by commercial smart devices,we theoretically analyze and verify the model of the acoustic absorption effect of different reflective surfaces.We find that the principle of signal amplitude and phase variations during the eye blink process.We propose a viewing position scheme to eliminate multiple interferences and amplify the signal variations caused by subtle movements,achieving a breakthrough in the sensing granularity based on acoustic signals.(3)To address the weak reflected signals in long-range sensing,we prove the impact of the number of mixing samples on the target reflected signal through theoretical analysis,and propose a virtual transceiver mechanism to improve the signal-to-noise ratio of the reflected signal.We propose a searching algorithm to detect the target in the environment and distinguish it from other interfering objects.The proposed method improves the sensing range of acoustic sensing.(4)To address the problem in large-scale sensing due to severe attenuation of acoustic signals at high frequencies,we propose a solution to deploy sensing devices on mobile platforms.For the large interference caused by the movement of the device itself,we propose a method to cancel the movement of the device itself,which employs a static object in the environment as a reference to obtain the movement information of the device itself and then subtract if from the target reflected signals.It can realize the sensing of human movement under the device movement scenario,further extending the sensing scale of the acoustic sensing.