Using a synergetic computer in speech recognition and identification of persons

The Synergetic Computer is a new computer concept, which can be used in different pattern recognition tasks. Based on the Synergetic Computer we develop an optic and an acoustic system for automatic recognition of spoken words. The optic speech recognition system uses the lip motions observed by a camera, the acoustic system uses the digitized sound signal measured by a microphone. Both systems consist of an appropriate preprocessing and the Synergetic Computer completing the actual classification task. In the optic preprocessing we obtain a feature vector from the image sequence by computing the optical flow. In the acoustic preprocessing we Short Time Fourier transform the sound signal to get a feature vector. The feature vectors characterize the lip motions or respectively the sound signal and serve as prototype or test vectors for the Synergetic Computer. As auditory and visual information are independent, more acurate and robust speech recognition can be achieved by combining the optic and acoustic system. Both the optic and acoustic system are not only able to recognize the spoken words but also the speaker. So both systems can also be applied to identify persons by the way they speak. In our experiments the system were able to distinguish 10 different words or 5 different persons.