5R-9
Audio-visual musical instrument recognition
○Angelica Lim(京大),中村圭佑,中臺一博(ホンダ Research Institute Japan),尾形哲也,奥乃 博(京大)
Is this person playing a violin or a flute? Classification of musical instrument performances is usually carried out using audio features such as spectral coefficients. We propose augmenting the typical audio feature set with visual features. We show that a combination of audio features and video perform better than audio alone, and verify this multimodal recognition approach on a real-time robot platform.