An Interactive Differential Evolution Method With Human Auditory Perception for Sound Composition

Interactive evolutionary computation (IEC) finds diverse applications in the domain of sound. However, there is a lack of methods and operations for combining sound elements into a composite sound during the crossover phase of sound composition. To this end, we propose a novel composite crossover operation using paired comparison-based interactive differential evolution. This operation integrates the target and mutant vectors into a composite sound through linear rescaling, enabling users to synthesize and assess their preferred sounds during the evaluation phase. Moreover, an optimization-stopping mechanism is incorporated to allow users to halt the process when a satisfactory sound is produced. This helps to alleviate user fatigue by eliminating unnecessary candidate sounds and improving the efficiency of the composition process by reducing redundant generations and time consumption. The efficacy of this operation was demonstrated through comparative testing and sound quality analysis. Furthermore, this article presents an original analytical approach based on four fundamental attributes of sound for analyzing human auditory preferences in composite sounds. This method combines both quantitative and qualitative paradigms, achieved through principal component analysis and human subjective auditory analysis. These discoveries have significant implications for both sound composition and the analysis of human auditory perception.