<?xml version="1.0" standalone="yes"?>
<Paper uid="H91-1031">
  <Title>Signal Representation Attribute Extraction and the Use Distinctive Features for Phonetic Classification 1</Title>
  <Section position="2" start_page="0" end_page="0" type="intro">
    <SectionTitle>
INTRODUCTION
</SectionTitle>
    <Paragraph position="0"> The overall goal of our study is to explore the use of distinctive features for automatic speech recognition. Distinctive features are a set of properties that linguists use to classify phonemes \[1,13\]. More precisely, a feature is a minimal unit which distinguishes two maximally-close phonemes; for example/b/and/p/are distinguished by the feature \[voicE\].</Paragraph>
    <Paragraph position="1"> Sounds are more often confused in relation to the number of features they share, and it is believed that around 15 to 20 distinctive features are sufficient to account for phonemes in all languages of the world. Moreover~ the values of these features, such as \[+HIGH\] or \[-aOUND\], correspond directly to contextual variability and coarticulatory phenomena, and often manifest themselves as well-defined acoustic correlates in the speech signal \[3\]. The compactness and descriptive power of distinctive features may enable us to describe contextual influence more parsimoniously, and thus to make more effective use of available training data.</Paragraph>
    <Paragraph position="2">  In order to fully assess the utility of this linguistically well-motivated set of units, several important issues must be addressed. First, is there a particular spectral representation that is preferred over others? Second, should we use the spectral representation directly for phoneme/feature classification, or should we instead extract and use acoustic correlates? Finally, does the introduction of an intermediate feature-based representation between the signal and the lexicon offer performance advantages? We have chosen to answer these questions by performing a set of phoneme classification experiments in which conditional variables are systematically varied. The usefulness of one condition over another is inferred fl'om the performance of the classifier.</Paragraph>
    <Paragraph position="3"> In this paper, we will report our study on the three questions that we posed earlier. First, we will report our comparative study on signal representations. Based on these results, we will then describe our experiments and results on acoustic attribute extraction, and the use of distinctive features. Finally, we will discuss the implications and make some tentative conclusions.</Paragraph>
  </Section>
class="xml-element"></Paper>