Establishment of a sound field reproduction method with high flexibility

具有高灵活性的三维声场的记录和再现对于远程协作系统中的高性能协作非常重要。本研究的目的是建立一种补偿双耳信号个性的声场再现方法，将双耳信号用两个麦克风嵌入假头（声学人体模型）的两只耳朵中进行录音，并通过耳机进行再现。本研究项目涉及两种声场：一种是2005财年的死场，另一种是2006财年的混响场。一个房间的混响是通过在房间中测量的声学传递函数的峰值和低谷来观察的。当ARMA（自回归移动平均）模型应用于传递函数时，可以通过控制z平面上极点和零点的数量和位置来控制混响。在2005财年，通过减少极点和零的数目平滑传递函数，得到了本研究提出的补偿函数。在2006财年，通过将ARMA模型得到的波峰和低谷相加，得到了混响室的补偿函数。通过心理声学实验对这些补偿功能的作用进行了评价。实验结果表明，在极点和零点数量足够的情况下，所提出的补偿函数对音质和声音定位都是有效的。除上述研究外，还进行了以下两个研究项目：1.研究对象；结果表明，在远程协作系统中，控制声场信息可以提高语音的可理解性。利用DSP（数字信号处理器）入门套件构建了双通道HpTFs（耳机到耳道传递函数）测量系统。

Recording and reproduction of a three-dimensional sound field with high flexibility is important for high-performance collaboration in a tele-collaboration system. The purpose of this research is to establish a sound field reproduction method that compensates for individuality of binaural signals, which are recorded with two microphones embedded in the two ears of a dummy head (acoustical mannequin) and reproduced via headphones.Two kinds of sound fields were addressed in the term of this research project : One was a dead field intended in FY 2005 and the other was a reverberant field intended in FY 2006. Reverberation of a room is observed as peaks and dips in an acoustic transfer function measured in the room. When the ARMA (Auto-Regressive Moving-Average) model is applied to the transfer function, reverberation can be controlled by controlling the numbers and locations of poles and zeros in the z-plane. In FY 2005, a compensation function proposed in this research was obtained by smoothing the transfer functions with reduced numbers of poles and zeros. In FY 2006, a compensation function for a reverberant room was obtained by adding peaks and dips which were derived by using the ARMA model. The effects of these compensation functions were evaluated by psychoacoustical experiments. Experimental results show that the proposed compensation functions are effective for sound quality as well as sound localization if the numbers of poles and zeros are adequate.Besides the above mentioned research, the following two research items were undertaken :1. It was shown that controlling sound field information in a tele-collaboration system brings high word intelligibility irrespective of talking face images.2. A two-channel measurement system of HpTFs (Headphone-to-ear-canal Transfer Functions) was constructed using a DSP (Digital Signal Processor) starter kit.