Face recognition with the synergetic computer

Synergetic Computers are a new approach for recognition and identification of patterns and objects. This paper presents results gained by applying software-simulated Synergetic Computers to the face recognition problem. The potential of Synergetic Computers for face recognition was already recognized by Haken, although his experiments took place under unrealistic conditions, e. g. no changes in face expression, viewing angle, etc. were allowed. To thoroughly investigate the potential of Synergetic Computers, 1000 face images of 20 persons were used, taken under different viewing angles varying in the range of several degrees. We also allowed changes of facial expressions, like grief, laughter, grimaces etc. Classification rates of 100 per cent were achieved on the learning-set and 98.8 per cent on the testset. The total learning time was about 10 minutes on a SPARC 10/40 workstation. With a newly developed class of Synergetic Computer algorithms, the number of required prototypes for c lassification could be strongly reduced without loss of relevant information. Compared to a Nearest Neighbour approach, classification speed was about 50 times better with this new algorithm, while classification rates were about the same. The time needed for the recognition of a single face lies within fractions of a second.