Abstract: In CASA（Computational Auditory Scene Analysis）, the main factor of source localization is ITD. To estimate ITD, conventional methods use cross-correlation of binaural firing rates of neurons. However, these methods have two obvious flaws with complex calculation and weak anti-noise ability, bringing a huge bottleneck to realize moving source real-time localization. The very prevalent method of Zero-Crossing Time Differences（ZCTD） is improved in the proposed method. The novel method uses Teager Energy Operator（TEO） to directly extract reliable ITD samples, and omits the procedure of extracting the reliable ITD samples by estimated SNR. The simulation results show that this method can more rapidly and more accurately estimate ITD and has strong anti-noise ability. This method is applied to locating the moving sound. This paper presents the model of moving sound location on zero-crossing points binaural time differences, and verifies the possibility of realizing moving sound real-time localization by simulation.

Key words: Computational Auditory Scene Analysis（CASA）, moving sound localization, zero-crossing points, Teager Energy Operator, Interaural Time Differences（ITD）

摘要： 人耳听觉定位的一个主要因素是ITD。在传统的计算声场景分析方法中，对ITD的估计采用对双耳神经发放率做互相关求得。但该方法具有计算量大和抗噪声能力弱两大缺陷，给实现运动声源实时定位带来了巨大瓶颈。对较新的过零时间差（ZCTD）方法进行了改进，采用Teager能量算子（TEO）提取语音包络截取可靠语音段估计ITD，避免了ZCTD方法中用信噪比的估值提取可靠ITD的繁琐步骤，能够更快速、准确地估计ITD，具有较强的抗噪声能力。将该方法运用到运动声源定位上，提出了基于过零点双耳时间差的运动声源定位模型，通过仿真实验验证了其实现运动声源实时定位的可行性。

关键词: 计算声场景分析, 运动声源定位, 过零点, Teager能量算子, 双耳时间差