UBC Theses and Dissertations
Towards spoken English : a computer based synthesizer for a reading machine for the blind Yeung, James Man Chung
A speech synthesizer has been developed to synthesize real-time speech directly from letter information. The synthesizer was implemented on a PDP-12 digital computer and its peripheral devices. Synthetic speech was produced by the concatenation of 33 basic phonemes. These phonemes were processed by a segmentation program which was developed to extract, record, display and store them: the original speech samples were treated similarly. The letter information from the output of a letter recognizer was stored in an input stack (or buffer) until a complete word has been received. Then the input word was synthesized by a synthesis program which assigned phonemes to the word according to the stored dictionary and the general grapheme-phoneme correspondence rules written in the program. A dictionary containing words that were not pronounced according to the above general rules, and in addition, a list of phoneme equivalents, were stored in a disk file system. The general rules were set up according to the most frequent grapheme-phoneme transcriptions and a frequency study of the first 1,000 words of the Thorndike word list. The synthesizer was capable of synthesizing over 90% of spoken English speech accurately and at the same time could be understood by the listener after a short period of training in spite of the mechanical sounding dialect which resulted from the incorrect pronunciation of some words. An intelligibility test using isolated words, sentences, and short passages generated by the synthesizer, was given to six subjects. The results showed that over 90% of the isolated words, 95% of the sentences and 98% of the short passages could be recognized by the test subjects after approximately one hour of training.
Item Citations and Data