Domain Adaptation for Mobile Device Acoustic Based Proximity Sensor

In our modern days, almost every person in the western world owns a relatively new smartphone. Every smartphone is equipped with an infra-red photoelectric proximity sensor, which is placed next to the phone’s speaker, and used for turning off the smartphone’s screen when the sensor is blocked. In example, when one is receiving a phone call, the person places the smartphone besides his ear, thus create a blockage on the proximity sensor that will turn off the screen.
We propose a different approach for creating a proximity sensor, which does not require special sensor, using an acoustic based proximity sensor – convenient method that does not require any additional hardware.
Our system utilizes the smartphone’s speaker and two microphones, the main microphone and the noise canceling microphone, and uses the microphones relationship to identify whether the smartphone’s screen is being blocked or not.
To identify the blockage, we use an SVM classifier on the recordings, that will label each sample as ‘near’ or ‘far’. Our algorithm uses the SVM trained once by one smartphone and a parallel transport method the create a domain adaptation that will allow us to use the same SVM for different mobile phones.
For our solution, we used a Xiaomi Mi 10 mobile phone as the main training model, Xiaomi Note 4 as a tested model and Samsung Note 5 as the verification model.
In this work we review the available method and its potential application and discuss future perspective of enhancing this technique.

Audio based classifier