9.7 Separating a moving sound.


Use this section if you do not know how to set UPDATE_METHOD_TF_CONJ and UPDATE_METHOD_W as properties of GHDSS .


If you do not understand well, use the default value. When changing it, you must determine if you are focusing on the sound source itself (ID) or the direction (POS). In particular, when a sound source moves or a robot’s body moves, separation results change under conditions in which good separation traceability is needed. When focusing on ID, reuse the ID from an earlier step as a geometric constraint and separation matrix value. Therefore, when a sound source moves fast, those values will become inappropriate and separation will be unsuccessful. In contrast, when a sound source does not move much, the separation accuracy will continue to improve during reuse. When focusing on POS, the characteristics are reversed compared with ID. A focus on POS is suited for when a sound source moves at high speed. When the interval between movements is large, a database is used for values of a constraint condition, and the separation matrix or system is initialized. An ideal way is to apply these constraints to each sound source dynamically. For details, see the description of the GHDSS module.



See Also