AUTOMATED OPTIMIZATION OF PARAMETERS FOR FM SOUND SYNTHESIS WITH GENETIC ALGORITHMS
Yuyo Lai, Shyh‐Kang Jeng, Der-Tzung Liu, Yo-Chung Liu · 2006
We propose a method to automate the optimization of the parameters of a FM (frequency modulation) synthesizer. Our goal is to find a set of parameters that can generate sounds similar to the target sound. By the Genetic Algorithm (GA), which is one type of evolutionary algorithms, we can find the optimal parameters automatically. The input is a wave format file as the target sound. A set of parameters is generated by the GA core, and the GA process starts. Parameters are evaluated for the survivability in the next generation. The fitness value of each parameter is assigned by processing spectral features. The selection step and variation step choose better parameters and form new parameters for the next generation. After several generations, the parameters will be more close to the solution that generates a similar sound to the target. The originality of this study is that we combine a timbral feature, the spectral centroid, and the spectral norm as a fitness value. Experiments are conducted and the results are rather satisfactory. 1.