HRI-JP HOME > Research > Publications > ロボット聴覚ソフトウェアHARKにおける音源定位パラメータチューニングの検討

December 2011
This paper addresses parameter tuning for Sound Source Localization (SSL) in open source robot audition software HARK. In a daily environment, there exist a lot of dynamically-changing sound sources. SSL for such sound sources are a difficult task, and thus parameter tuning is necessary. However, not a few parameters mutually affect SSL performance. We proposed a hybrid approach of automatic and manual parameter tuning to solve this problem. It provides a simple and highly-usable interface for tuning. Our case studies showed that it takes only 15 min for a tuning task with the proposed method, while an expert does 2-3 days with fully-manual tuning.