Multiple packet reception based on the time-varying transmitted power form of implicit training

We propose a new algorithm for the multipacket reception task. In our proposal, the transmitters change their output power in a symbol by symbol fashion. To achieve packet separation, we exploit the time-varying power variation of the sources. The separation solution is obtained in closed form and it is shown that perfect interference cancellation is asymptotically obtained under noiseless conditions. Furthermore, when noise is present, the method asymptotically approaches the optimum minimum mean square error solution. Simulations results are presented that corroborate the theory and illustrate the performance of the proposed approach for finite length data packets.