Analogue radial basis function networks for phoneme recognition
Edward Gatt, Joseph Micallef · 2003
This paper presents an analogue radial basis function neural network for phoneme recognition. The neural network has been implemented on-chip using 0.35 /spl mu/m three-metal dual-poly CMOS technology. Radial basis function neural networks have been adopted because they offer improved training times when compared to multi-layer perceptron networks implementing conventional back-propagation learning (S. Renals and R. Rohwer, Proc. IEEE/INNS First Inter. Joint Conf. Neural Networks, vol. 1, pp. 461-467, 1997). The paper also presents the performance characteristics for the chip, together with its application to the problem of phoneme recognition.