Investigations on active perception in speech production and hearing systems

精确的源信息提取程序被发明出来，为进一步的调查提供了不可或缺的基础。主要研究目标是根据首席研究员在以前的研究中的发现和Kawato等人提出的“运动控制的一般理论”，开发F0（基频）控制模型，使用新程序获得的数据产生。该模型的一个工作原型，一个scat渲染模型，是基于一个非常高质量的语音分析，修改和合成方法，直，由首席研究员开发。这些成就的一部分在美国声学学会第141次会议上进行了介绍，给与会者留下了深刻的印象。这一影响导致邀请其作为特邀发言者参加将于2002年6月举行的第143次会议。新的源信息提取程序（用于频域、时域和非周期性属性）也被应用于开发语音质量分析和再合成的程序。这些程序是在意大利举行的国际会议MAVEBA上介绍的，也产生了很大的影响。在丹麦举行的2002年EUROSPEECH会议上提出的一篇论文是一个典型的例子。它表明，新的程序是足够精确的，以检测系统的F0毛刺周围的辅音-元音和元音-辅音过渡，也足够强大的渗透到现象的起源。尽管事实上，毛刺不是由感知和生产之间的积极相互作用，发现和调查具有科学意义，并说明了有用的程序。最后，研究结果进一步表明，一个新的，仍然模糊的，听觉信息表示的基础上固定点的概念将出现作为一个实例的广义听觉处理原则。

Precise procedures for source information extraction were invented for providing an indispensable basis for further investigations. The major research goal, development of an F0 (fundamental frequency) control model based on an integration of findings of the head investigator in his previous investigations and "general theory of motor control" proposed by Kawato et.al., was yielded using data acquired by the new procedures. A working prototype of the model, a scat rendering model, was implemented based on a very high-quality speech analysis, modification and synthesis method, STRAIGHT, developed by the head investigator. A part of these accomplishments was presented at the 141st meeting of the Acoustical Society of America and gave a strong impression on the participants. The impact resulted in an invitation to the 143rd meeting, which will be held in June 2002, as an invited talker.The new source information extraction procedures (for frequency domain, time domain and aperiodicity attributes) were also applied for developing procedures for voice quality analysis and resynthesis. The procedures were introduced at MAVEBA, an international conference held in Italy and also gave a strong impact. A typical example to demonstrate the power of the procedures is a paper presented at EUROSPEECH'2002 in Denmark. It illustrated that the new procedures are precise enough to detect systematic F0 glitches around consonant-vowel and vowel-consonant transitions and also powerful enough to penetrate into the origin of the phenomenon. In spite of the fact that the glitches are not caused by active interactions between perception and production, the finding and investigations have scientific importance and illustrates the usefulness of the procedures.Finally, the findings further suggest that a new, still vague, auditory information representation based on fixed-point concept will be emerged as an instantiation of a generalized auditory processing principle.