Spelled Speech as an Output for Computers and Reading Machines for the Blind

The development of a speech display system for the blind built around the TX-0 and PDP-1 computers at the Massachusetts Institute of Technology is described. The system, dubbed ``SPANC,'' was designed to permit the application of a variety of sound compression techniques to be used. Spelled-speech alphabets and displays were developed. An investigation of the letter sounds was carried out for the purpose of revealing the positions where cuts could be made so as to achieve high reading rates yet retain smooth coalescence. Identification of the positions of these cuts could be made with gross features of the acoustic waveform. Cuts of more than 50-ms duration could be made in the region of the articulatory transitions for the consonants, and very acceptable representations of the letter sounds could be obtained by preserving this region for the vowels. Finally, the overlapping of letter utterances as a compression technique was accomplished with subjects experiencing very little difficulty for overlaps corresponding to reading rates of 90 and 120 words per minute.