This paper proposes a novel particle swarm optimization algorithm based on sexual reproduction for speech enhancement. In the proposed algorithm, the population is divided in two equal subgroups: male and female. The male subgroup pays more attention to the global search, but the female subgroup focuses on the local search. Therefore, females utilize local search capability in prosperous regions, while males search the problem space globally to avoid entrapment in a local minimum. The crossover operation between two groups adds new solutions to the population. Experimental results indicate that the sexual reproduction-based particle swarm optimization (SRPSO) outperforms the standard particle swarm optimization (SPSO), and gradient-based NLMS algorithm in speech enhancement application.