GMSK demodulation using K-means clustering techniques

Recent applications of clustering and neural network techniques to channel equalization have revealed the classification nature of this problem. This paper illustrates an implementation of a GSM receiver in which channel equalization and demodulation are realized by means of the K-means algorithm. The advantage is found in the significant reduction of the computational complexity with respect to the classical MLSE equalizer. The performance of the proposed receiver, evaluated through a channel simulator for mobile radio communications, are compared with results obtained by means of a classical 16 states Viterbi processor. It is shown that despite the strong simplification in the receiver, performances are still acceptable considering the ETSI parameters.