Simplified speech recognition techniques for microcomputers

Speech recognition systems are becoming popular devices for improving communication between man and machines. The task of speech recognition is very difficult, resulting in poor accuracy, slow response time, limited vocabulary, and expensive hardware. The intent of this study was to produce a simplified method of speech recognition which would have a moderate level of performance yet would be inexpensive to In particular, a goal was set to have the algorithm run in real time on a microcomputer.

The resultant method is a software adaptation of a machine designed and patented by Donald Lokerson.¹ The program runs on an Apple II computer equipped with an Intel 8088 microprocessor and with an analog-to-digital converter. The speaker-dependent program can distinguish among ten words at a 90% success rate while requiring minimal delays for processing.