Quasi-Opposition-Based Learning in a Shuffled Particle Swarm Optimization for Solving Frequency Modulation Sounds Parameter Identification Problem

This paper proposes some novel versions of the shuffled particle swarm optimization (SPSO) for solving the frequency modulation sound parameter identification (FMSPI) problem. In the SPSO, a population is divided into several parallel groups and then each group is independently evolved in an evolutionary process using a particle swarm optimization (PSO). This paper employs two different strategies to prevent a premature convergence and providing a better balance between the exploration and exploitation abilities of the SPSO algorithm. Firstly, it proposes that we can use a separate strategy for the inertia weight factor parameter of each group in each iteration of the SPSO algorithm. For the second strategy to provide a deep search of promising areas, a quasi-opposition-based strategy is inserted in the SPSO. Experimental results on FMSPI problems show that new employed strategies reduction lead to achieving a more effective and robust algorithm so as it can considerably improve the performance of the SPSO.