An Investigation of Locus Equations as a Source of Relational Invariance in Speech

摘要言语不变性的实验和理论追求涉及到语言抽象的多个层面。这项研究的目的是研究“轨迹方程”算法作为音调发音位置关系不变性的可能来源。在初步工作中，轨迹方程-回归到由(I)CV转换开始的频率和(Ii)元音目标定义的一组坐标-的特征是区别于元音的停止位置的大小。计划在跨语言和不同学科人群中继续开展这项工作。初步发现的佐证将有力地暗示更高级别的抽象的可用性，可以有效地编码停留地点分类的关系不变性。共振峰转换起始点和偏移点的三维图形投影，与统计线性判别函数分析相结合，将检验作为音位分类组织框架的“充分知觉对比”的概念。这项研究的意义在于加深我们对语音音素水平解码的理解，并最终更好地概念化大脑如何“光谱”地代表音素等值类别。

ABSTRACT The experimental and theoretical pursuit of invariance in speech has involved many levels of linguistic abstraction. The objective of the research is to investigate the "locus-equation" algorithm as a possible source of relational invariance for place of articulation in stops. Locus-equations- regression fits to a set of coordinates defined by the frequency of (i) the onset of CV transition, and (ii) the vowel target- were shown, in preliminary work, to be characterized by distinctive slopes differentiating the stop place dimension independently of the vowel. Continuation of this work cross-linguistically, and across a diverse subject population is planned. Corroboration of preliminary findings would strongly imply the availability of a higher order level of abstraction that could effectively code relational invariance for stop place categorization. Three- dimensional graphic projections of formant transition onset and offset points, in combination with statistical linear discriminant function analyses, will test the notion of "sufficient perceptual contrast" as an organizational framework for phonemic categorization. The significance of this research is to further our understanding of phonemic-level decoding of speech, and ultimately, to better conceptualize how the brain might "spectrotopically" represent phonemic equivalence classes.