FM聲音合成器系統之最佳參數設計
Automated Optimization of Parameters for FM Sound Synthesis with Genetic Algorithms
Date Issued
2006
Date
2006
Author(s)
Lai, Yu-Yo
DOI
en-US
Abstract
In this thesis we propose a system to optimize the parameters of an FM (Frequency Modulation) synthesizer. The system can find the parameters of FM synthesizers based on GA (Genetic Algorithm), and generate a sound similar to the target sound. In the GA process, the parameters of the FM synthesizer are viewed as genes, and the set of parameters that can generate a sound is viewed as an individual. Initially, the system inputs a wave format file as the target sound, and then the GA core randomly initializes a set of parameters for FM synthesis. Then, the system calculates the fitness value for each individual and selects individuals for the variation step through natural selection. In the variation step, different operations are applied to modify the genes of the individuals. Finally, the process moves on to the next generation. After several generations, the individuals will be close to the solutions and contain proper parameters to generate similar sounds. In this thesis, we also propose a novel approach that combines two timbral features: spectral norm and spectral centroid distance as the fitness function of GA process. We conducted experiments with five target tones: grand piano, celesta, organ, clarinet and nylon guitar. Results of the experiments show that our approach outperforms the conventional approach that only applied spectral norm. We also conducted experiments for deciding the weighting of the combined features and determining the characteristics of selected timbral features.
Subjects
FM合成
基因演算法
音色特徵
參數最佳化
FM Synthesis
Genetic Algorithm
Timbral Features
Optimization for Parameters
Type
thesis
File(s)![Thumbnail Image]()
Loading...
Name
ntu-95-R93942114-1.pdf
Size
23.31 KB
Format
Adobe PDF
Checksum
(MD5):8acddf9ad26f3d97514221a59bef7cbe