Real-time High-precision Motion Tracking Based On Commercial Earphones

For most smart applications,tracking people's motion in real time is crucial.Among the many existing solutions,the acoustic-based has unique advantages due to low cost and high accuracy.Because of the ubiquity of smart phones,most of the exist-ing acoustic-based motion tracking systems are implemented using the built-in speaker and microphone of smartphones.However,the built-in microphone and speaker are usually fixed on the body and the position cannot be changed freely,which makes the application scenarios limited greatly.In this dissertation,we focus on how to use com-mercial earphones to improve the flexibility and mobility of current acoustic tracking,so that acoustic-based tracking can trigger more novel applications.The most innovative idea is to indirectly track the movement of the hand holding the earphone by track-ing the movement trajectory of the earphone in space,and transfer the captured move-ment to smart devices,for example,smartphones,connected to the earphone in real time,thereby realizing a new kind of human-computer interaction.Before we turn the straightforward idea into a practical system,we need to tackle three challenges.First,a strong self-interference signal exists in the wired earphone,which masks the sound wave signal from the air received by our microphone.Second,there is a common frequency offset in Bluetooth earphones,which causes the frequency difference between the ac-tually generated sound wave signal and the expected signal and results in inaccurate motion tracking.Third,the available inaudible bandwidth of the Bluetooth earphone is narrower,which makes it impossible for us to use high bandwidth to achieve very high tracking accuracy.This dissertation proposes feasible technical solutions for these challenges.First of all,this dissertation carefully studies the cause of self-interference signal,builds a mapping model between the self-interference signal and the transmit-ted signal,and proposes two self-interference cancellation solutions based on Hilbert transform and adaptive filtering.Then,this dissertation introduces an error correction method to compensate the distance deviation caused by frequency offset.Finally,for limited available inaudible bandwidth,this dissertation presents a fine-grained distance measurement scheme based on the number of sine wave peaks and valleys,which oc-cupies the most narrow bandwidth to achieve high-precision motion tracking.After overcoming the above three challenges,this dissertation implements and evaluates ID,2D and 3D tracking systems on Android and Linux PC platform using commercial ear-phones.Experimental data shows that our system can achieve high-precision real-time motion tracking while ensuring flexibility.