Biological systems perform pattern recognition using interconnections of large numbers of cells called neurons. The large number of parallel neural connections makes the human information processing system adaptable, context-sensitive, error-tolerant, large in memory capacity, and real-time responsive. These characteristics of the human brain provide an alternative model to the more common serial, single-processor signal processing architecture. Although each human neuron is relatively slow in processing information (on the order of milliseconds), the overall processing of information in the human brain is completed in a few hundred milliseconds. The processing speed of the human brain suggests that biological computation involves a small number of serial steps, each massively parallel. Artificial neural networks attempt to mimic the perceptual or cognitive power of humans using the parallel-processing paradigm. Table 7.1 compares the features of artificial neural networks and the more conventional von Neumann serial signal processing architecture.
Online access to SPIE eBooks is limited to subscribing institutions.