唱歌機器人音效之DSP實現

Abstract

在卡拉ok機與多媒體系統中，音效往往伴演著舉足輕重的角色；現今在許多3C產品上，亦可看見許多音效的應用，例如卡拉ok機中的迴響(Echo)，電話及電視媒體常使用的變聲器(Pitch-shifter)等等。早期的各種音效主要是利用訊號延遲的原理去產生，而後隨著各種處理器運算速度的提升，現今的音效不但要求娛樂性高，擬真性好，同時也需要運算量少這方面的考量。在本論文中，主要在探討目前卡拉ok娛樂系統上常看見的各種音樂特殊效果，從物理原理及數學方程式上出發，將之實現至CPU與一特定DSP內，並試著在運算量上最佳化，成為較具新意的各種音效演算法。此外，將這些音效實現在娛樂型機器人上。

Audio effects, such as delay, echo, flanging, chorusing, pitch shifting, vibrato, tremolo and etc…, and virtual bass are indispensable in music productions and performance. Some are also available for home entertainments and car audio systems. Most of these effects are implemented using digital signal processors, which may reside in separate modules or may be built into keyboard workstations and tone generators. In this thesis, DSP implementation of various digital audio algorithms will be demonstrated, with theoretical equations as well as actual programming implementations shown wherever possible. These include basic audio signal manipulation, filter techniques, waveform synthesis techniques, digital audio effects theories and so on. Finally, these audio effects will implement on a singing robot.