Estudo comparativo entre os algoritmos de processamento de áudio phase vocoder e spectral modeling synthesis

Abstract

This work presents a comparative analysis based in qualitative and quantitative criteria between the frequency domain audio processing algorithms Phase Vocoder and Spectral Modeling Synthesis (SMS). The algorithms were applied on synthesized waves in time-stretch and pitch-shift operations. Initially, the difference between the original and processed waves was calculated seeking evaluate quantitatively the impact of each algorithm. Later, new samples were selected for interest cases, like percussive instruments, for example. After the application of the algorithms, each sample was shown to volunteers, which evaluated it using a questionnaire. It was notable that the SMS' samples had a larger quantity of artifacts (defects) than Phase Vocoder's, receiving a worst evaluation, also generating noisier waves when applied in the synthesized waves. For future work, it's suggested using the SMS in it's capabilities of analysis and generation of effects in simpler sounds.