Sounds Source Segregation

描述（由申请人提供）：我们拥有一种非凡的能力，可以“听出”（分离）构成我们日常听力中遇到的复杂声学环境的单个声源。理解这种能力一直是心理声学研究的重点，但这是一个巨大的挑战。摄动分析是一种方法学方法，它对视觉相关问题有深入的了解[Murray， R.F. 2011]。[j].中国生物医学工程学报，2016,33(2):444 - 444。在这里，这种方法适用于试听。应用分三个阶段进行：首先，根据发声源的生成模型合成简单的语音和环境声音。其次，从非目标（噪声）源中分离目标的听者决策策略由生成模型决定，由听者对目标的判断与声学参数的合法扰动相关的决策权重（回归系数）决定。第三，通过将获得的权重和残差与最大似然（ML）观测器的权重和残差进行比较，确定限制偏析的因素，最大似然（ML）观测器根据产生源的运动方程优化偏析。该方法代表了传统方法的重大进步，传统方法通过表演准确性或在非法反对中放置声学线索的效果来推断听者的决策策略。具体目的是将该方法应用于(1)测试两类主要的分离模型，目标增强与噪声消除，(2)测试基于声学的与基于力学的关于源分离的假设的预测，共同命运与共同来源，(3)量化目标噪声不确定性和相似性的相对影响，(4)确定水平优势对声源分离的影响，(5)通过使用决策权重来识别个体倾听风格，以解释不同任务的不同行为模式。预计从这些研究中获得的知识将为声源隔离问题的更广泛方面的研究提供信息，包括与听力障碍的评估和治疗相关的研究，因为它影响了嘈杂环境中的日常听力。

DESCRIPTION (provided by applicant): We possess a remarkable ability to 'hear out' (segregate) the individual sound sources that makeup the complex, acoustic environments we encounter in everyday listening. Understanding this ability has been a preoccupation of psychoacoustic research, but it is a tremendous challenge. Perturbation analysis is a methodological approach that has advanced understanding of related problems in vision [Murray, R.F. 2011. J. of Vision 11, 1-25]. Here the approach is adapted to audition. The application proceeds in three stages: First, simple speech and environmental sounds are synthesized according to a generative model of the sound-producing source. Second, listener decision strategy in segregating a target from non-target (noise) source is determined from decision weights (regression coefficients) relating listener judgments regarding the target to lawful perturbations in acoustic parameters, as dictated by the generative model. Third, factors limiting segregation are identified by comparing the obtained weights and residuals to those of a maximum- likelihood (ML) observer that optimizes segregation based on the equations of motion of the generating source. The approach represents a significant advance over traditional methods that infer listener decision strategy from performance accuracy or from the effect of placing acoustic cues in unlawful opposition. Specific aims are to apply this method to (1) test between two major classes of segregation models, target enhancement vs. noise cancellation, (2) test predictions of acoustics-based versus mechanics-based hypotheses regarding source segregation, common fate vs. common source, (3) quantify the relative influence of target-noise uncertainty and similarity, (4) determine the influence of level dominance on sound source segregation, and (5) account for diverse patterns of behavior across tasks by using the decision weights to identify individual listening styles. It is expected that the knowledge gained from these studies will inform research on broader aspects of the problem of sound source segregation, including those relevant to the evaluation and treatment of disordered hearing as it impacts everyday listening in noisy environments.