Year: 2018
Type: dissertation
Abstract: In audio time-scale modification (TSM), the duration of an audio recording is changed while retaining its local frequency content. In this thesis, a novel phase vocoder based technique for TSM was developed, which is based on the new concept of fuzzy classification of points in the time-frequency representation of an input signal. The points in the time-frequency representation are classified into three signal classes: tonalness, noisiness, and transientness. The information from the classification is used to preserve the distinct nature of these components during modification. The quality of the proposed method was evaluated by means of a listening test. The proposed method scored slightly higher than a state-of-the-art academic TSM technique, and similarly as a commercial TSM software. The proposed method is suitable for high-quality TSM of a wide variety of audio and speech signals.
Source: Aaltodoc (Aalto University)
Author: Damskägg, Eero-Pekka
Language: English
Topic: Music and Audio Processing
Subfield: Signal Processing
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
Funder: Aalto-Yliopisto
OpenAlex ID: W7123372502