Abstract:
The paper presents an approach to the design of a real-time Arabic digits recognizer. The spoken digit is filtered and passed to a microcomputer via a suitable analogue-to-digital (A/D) converter. The feature to be extracted is the duration of the positive slopes between zero-crossings. A vector is constructed for each of the spoken digits. Each element of this vector represents the sum of the durations that have the same zero-crossing and positive slope intervals. The recognition rate achieved by this method is 97% for a single speaker.