基于智能手机TDOA估计的被动声源定位方法与系统实现

Passive acoustic source target positioning method based on smart phone platform TDOA estimation and system implementation

被动目标监测广泛应用于国防、安全等领域。针对被动声目标监测,首次提出了基于手机平台的被动式声目标定位方法,并设计和开发了原型系统。针对手机间时钟同步对被动式信号到达时间差精度的影响,采用手机内置的麦克风和扬声器发送和接收同步声信号的方式,从而避免了节点间无线同步到声信号采集间的时间延迟不确定性。为了提高信号发送和接收时刻的估计精度,将时间戳信息调制到设定的线性调频声信号,并采用广义互相关方法来实现信号波形检测。进一步,针对被动目标声源的非合作特性,采用统计判决理论和语音活动性检测相结合的方法对观测信号进行联合检波,获得被动目标声源到达时间的高精度估计,即被动声源到手机间的到达时间差。最后,通过多部手机搭建了原型系统并设计实验,结果表明定位误差不超过10%的概率达到80%。

Passive acoustic target monitoring is widely used in national defense, security and other fields. Aiming at passive acoustic target monitoring, a passive acoustic target positioning method based on universal smart phone platform is put forward for the first time, and a prototype system is designed and developed. Aiming at the influence of the clock synchronization among smart phones on the TDOA accuracy of the passive signals, the method of using the phone built-in microphone and speaker to transmit and receive synchroni- zation acoustic signals is adopted, which avoids the time delay uncertainty from the inter-node wireless synchronization to the acoustic signal acquisition. In order to improve the estimation accuracy of signal transmission and receiving time, the time stamp information is modulated into the chirp acoustic signal set previously, and the generalized cross correlation （GCC） method is adopted to achieve the signal waveform detection. Furthermore, aiming at the non-cooperative characteristics of the passive target sound source, a method combining statistical decision theory and Voice Activity Detection （VAD） is adopted to detect the sound wave of the observed signal, achieve high-precision estimation of the arrival time for passive target sound source, i.e. the time difference of arrival （TDOA） between passive sound source and mobile phones. Finally, a prototype system was built based on multiple universal smart phones; experiments were designed and conducted, experiment results show that the probability for the positioning error not greater than 10% reaches 80%.