A SYSTEM FOR DATA-DRIVEN CONCATENATIVE SOUND SYNTHESIS

摘要: In speech synthesis, concatenative data-driven synthesis methods prevail. They use a database of recorded and unit selection algorithm that selects the segments match best utterance to be synthesized. Transferring these ideas musical sound allows new method high quality synthesis. Usual are based on model signal. It is very difficult build would preserve entire fine details sound. Concatenative achieves this by using actual recordings. This approach (as opposed rule-based approach) takes advantage information contained in many For example, naturally sounding transitions can synthesized, since aware context units. The CATERPILLAR software system has been developed allow high-quality instrument with level control, explorative free from arbitrary databases, or resynthesis recording sounds database. software-engineering concept component-oriented software, increasing flexibility facilitating reuse.