A new method for block demodulation without channel estimation

For communication channels exhibiting fading and interference, noncoherent maximum likelihood (ML) block demodulation generally requires the knowledge of received signal strength, fading levels, and the phase off-set introduced by the channel. Furthermore, ML block demodulation can be computationally prohibitive. This paper presents a sub-optimal block demodulation method based on a computationally simple and efficient clustering algorithm known as the K-means algorithm. The performance of noncoherent block demodulation based on this algorithm is determined for various phase and amplitude-phase modulation techniques. This demodulation scheme does not require knowledge of the received signal strength, fading levels, or channel phase offset. Numerical results show significant improvement over classical differential demodulation techniques.