生物启发下的机器人声源定位.pdf

Bio-inspired robot sound localization Seminar of Bio-inspired artificial intelligence Bio-inspired Sound Localization 1 Outline §  Motivation §  Reviews §  Model Architecture •  Auditory periphery model •  azimuth localization •  Ideal binary mask estimation •  Target signal reconstruction §  Evaluation §  Conclusion Bio-inspired Sound Localization 2 Motivation §  Cocktail Party Problem At a cocktail party, one can selectively attend to a single voice and filter out all the other acoustical interferences. How to simulate this perceptual ability on robot? Bio-inspired Sound Localization 3 Reviews §  Interaural acoustic cues: •  Interaural time difference (ITD) •  Interaural level difference (ILD) (/ interaural intensity difference(IID)) •  Integrated in the Inferior Colliculus Figure 2 Interaural time differences[1] §  Cross correlation algorithm •  Cross correlation ​ (Time frame, time lag) •  Cross correlogram C(frequency, azimuth, time lag) Bio-inspired Sound Localization 4 Model Architecture §  Model Overview Figure 3 Model Architecture §  Four stages (1)  model the auditory periphery; (2)  binaural cue extraction and azimuth localization; (3)  ideal binary mask estimation; (4)  Target signal reconstruction Bio-inspired Sound Localization 5 Psychoacoustic Model §  Auditory threshold of hearing •  Human hearable frequencies: 20 Hz to 20,000 Hz •  critical bandwidths §  Mask Effect “A