基于Kinect的动态虚拟听觉重放

发布时间：2018-02-09 15:42

本文关键词： 动态虚拟听觉重放头相关传输函数 Kinect 5.1通路环绕声　出处：《华南理工大学》2016年硕士论文　论文类型：学位论文

【摘要】：动态虚拟听觉重放是一种新的空间声重放技术。它头部跟踪器能实时监测倾听者的头部运动,而后根据空间信息的变化,实时调整信号处理,合成双耳信号及其随头部运动的动态变化,从而在耳机重放中产生相应的动态空间听觉事件。相比于传统的稳态虚拟听觉重放,动态重放包含了声源定位的动态因素,能够为倾听者带来更真实的听觉感知。动态虚拟听觉重放在双耳听觉的科学研究、虚拟现实、多媒体个人计算机与手持播放设备的文化娱乐方面有重要的应用价值。目前国内外已有不同的课题组发展了此类系统。这些系统很多是为科学研究和工程应用而发展,虽然其重放性能与精度可达到较高的要求,但其总体成本较高且使用操作较为复杂,并不适合一般个人计算机应用。另外,动态虚拟听觉重放需要用超高分辨率的头相关传输函数(HRTF)进行信号处理,受测量技术和测量工作量的限制,目前测量得到的HRTF数据的空间分辨率不能满足动态虚拟听觉重放的要求。为了解决这些问题,本文展开了以下相关工作:首先,研究了改善边界元计算HRTF稳定性的方法,并用该方法数值计算得到了KEMAR人工头模型的远场空间高密度HRTF数据,其空间分辨率达1°,频率上限达到18000Hz。然后,提出和建立了适用于多媒体个人计算机文化娱乐的动态虚拟听觉重放系统。系统的硬件系统包括Kinect体感摄像头作为头部跟踪器、个人计算机作为信号处理与控制系统,具有低延时特性的ASIO声卡作为声频信号输出。系统软件采用C++语言开发,包括三个功能模块,分别实现了人机交互、信号处理和输出、头跟踪三个功能。其中信号处理模块使用了基于主成分分析法的动态处理方法,提高信号处理效率同时避免了动态切换HRTF带来的噪声。跟踪模块使用了Kinect体感摄像头和微软的开发工具包,不但降低了成本,且使用操作简便。对所建立系统的动态性能客观参数进行了测量。测量的参数包括系统场景刷新率以及系统滞后延时。测试结果表明该系统能够满足动态重放的基本需求。最后基于本文所建立的动态重放系统,提出并实现了5.1通路环绕声的动态双耳重放,并采用虚拟声源定位实验进行了验证。实验结果表明,本文的动态系统可明显降低感知虚拟源的前后混乱率以及头中定位率。同时,系统对虚拟源的方位角定位准确性也有明显的改善,但对于仰角定位则改善并不明显。该实验结果也从一个侧面证明了本系统具有较好的虚拟听觉重放效果。
[Abstract]:Dynamic virtual auditory playback is a new spatial sound playback technique. The head tracker can monitor the head movement of the listener in real time, and then adjust the signal processing in real time according to the change of spatial information. The synthetic binaural signal and its dynamic changes with the head motion can produce the corresponding dynamic spatial auditory events in the headphone playback. Compared with the traditional steady-state virtual auditory replay, the dynamic playback includes the dynamic factors of the sound source location. Dynamic virtual hearing is placed in the scientific research of binaural hearing, virtual reality, Multimedia personal computers and handheld playback devices have important application value in culture and entertainment. At present, different research groups at home and abroad have developed such systems. Many of these systems have been developed for scientific research and engineering applications. Although its replay performance and precision can reach higher requirements, but its overall cost is high and the operation is more complex, so it is not suitable for general personal computer applications. Dynamic virtual auditory replay requires signal processing with ultra-high resolution header correlation transfer function (HRTF), which is limited by measurement techniques and measurement workload. The spatial resolution of the measured HRTF data can not meet the requirements of dynamic virtual auditory playback. In order to solve these problems, the following related work has been done in this paper: firstly, the method of improving the stability of BEM computing HRTF is studied. By using this method, the far-field high-density HRTF data of the KEMAR artificial head model are obtained. The spatial resolution is 1 掳and the upper frequency limit is 18000 Hz. A dynamic virtual auditory playback system suitable for multimedia personal computer culture and entertainment is proposed and established. The hardware system of the system includes Kinect camera as head tracker, personal computer as signal processing and control system. The ASIO sound card with low delay time is used as audio signal output. The system software is developed with C language, including three functional modules, which realize man-machine interaction, signal processing and output, respectively. The signal processing module uses a dynamic processing method based on principal component analysis. Improve the efficiency of signal processing and avoid the noise caused by dynamic switching HRTF. The tracking module uses Kinect camera and Microsoft development kit, which not only reduces the cost, but also reduces the cost. The objective parameters of dynamic performance of the established system are measured, including the refresh rate of the system scene and the delay time of the system. The test results show that the system can satisfy the dynamic replay. Basic requirements. Finally, based on the dynamic playback system established in this paper, The dynamic binaural replay of 5.1 channel surround sound is proposed and realized, and the virtual sound source localization experiment is used to verify it. The experimental results show that, The dynamic system in this paper can obviously reduce the rate of antecedent confusion and head localization of the perceived virtual source. At the same time, the accuracy of the azimuth positioning of the virtual source can also be improved obviously by the system. But the improvement of elevation location is not obvious. The experimental results also prove that the system has a good virtual auditory replay effect from one side.
【学位授予单位】：华南理工大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TN912.3;TN642

【相似文献】