This paper presents a novel membrane algorithm, called particle swarm optimization based on P systems (PSOPS), which combines P systems and particle swarm optimization. The PSOPS uses the representation of individuals, evolutionary rules of particle swarm optimization, and a hierarchical membrane structure and transformation or communication-like rules in P systems to design its algorithm. Experiments conducted on seven bench function optimization problems and time-frequency atom decomposition demonstrate the effectiveness and practicality of the introduced method.