Improved multipulse algorithm for speech coding by means of adaptive boltzmann annealing
Enzo Mumolo, Alessio Rebelli, Giuseppe Riceardi · European Transactions on Telecommunications · 1994
Abstract In this paper we will describe an algorithm, based on the Simulated Annealing technique, for solving the problem of multipulse excitation computation for speech coding applications. Multipulse excitation modeling is described by two set of parameters: pulse locations and amplitudes. The search for optimal pulse locations is the most burdensome task, since there is no closed form solution. Therefore, suboptimal approaches are used in order to maintain a good effectiveness in the parameter estimation while keeping the computational load within reasonable bounds. As compared to the one described in [2], the proposed algorithm achieves an effective improvement in objective performances of more than 1.5 dB. The price that must be paid for this performance improvement is the higher computational cost that this algorithm requires. The objective results of the proposed algorithm will be compared with the results of the algorithm described in [2], and the computational complexity will be analyzed.