Sound Source Localization for Enhancement of Orchestral Music for Multi-Sensor Recordings

TitleSound Source Localization for Enhancement of Orchestral Music for Multi-Sensor Recordings
Publication TypeMaster Thesis
Year of Publication2015
AuthorsLizarraga, X.
AbstractThis research is focus on the combination of Sound Source Localization (SSL) methods with Sound Source Separation (SSS) techniques inside an Orchestral music’s framework. The position at each source is estimated by a 3D grid search algorithm, by means Generalized Cross-Correlation-PHAse Transform (GCC-PHAT) feature vectors and the distances between microphones. This approach could be useful for audio applications in real time as automatic camera steering, videogaming, autonomous robots or auralization techniques. The main scope of this work is to learn in depth into signal processing algorithms for SSL, separation and audio enhancement with microphone arrays in a multi-track music recording. So, supplying a robust and low-complexity method for music applications as the up-mixing of the acoustic scene in other formats (mono-to-stereo, 5.1, Dolby Digital or other) or an audio rendering motor able to surfing through the acoustic scene. Some studies are done to combine SSL with SSS, nevertheless, most of them propose experiments with speech signals. Therefore, in this dissertation we have performed some experiments to assessing, by means of objective metrics, the combination of these processes with Orchestral music.
intranet