A Study of the Building Blocks in Speech
Cyril M. Harris · The Journal of the Acoustical Society of America · 1953
Identification of the information-bearing elements of speech is important in applying recent thinking on information theory to speech communication. One way to study this problem is to select groups of building blocks and use them to form standardized speech which then may be evaluated; a method having the advantage of simplicity is described. Individual recordings of the building blocks were made on magnetic tape and then various pieces of tape were joined together to form words. Experiments indicated that speech based upon one building block for each vowel and consonant not only sounds unnatural but is mostly unintelligible because the influences on vowel and consonants are missing which ordinarily occur between adjacent speech sounds. To synthesize speech with reasonable naturalness, the influence factor should be included. Here these influences can be approximated by employing more than one building block to represent each linguistic element and by selecting these blocks properly, taking into account the spectral characteristics of adjacent sounds so as to approximate the time pattern of the formant structure occurring in ordinary speech. There is no a priori method of determining how many building blocks are required to produce intelligible standardized speech. This can only be determined from experiments involving listening tests. Such tests are described.