A robust audio searching method for cellular-phone-based music information retrieval

We propose a search method for detecting a query audio signal fragment in long audio recordings. The query signal is assumed to be captured by a portable terminal, such as a cellular phone, in the real world. A major problem in this kind of search is that the features of the query sound may include distortions due to terminal characteristics or environment noise. The method proposed comprises local time-frequency-region normalization and robust subspace spanning. The former is used to make features invariant to additive noise and frequency characteristics, and the latter to choose frequency bands that minimize the effect of feature distortions. Experiments using cellular phones in the real world show the proposed method is effective.

[1]  Kunio Kashino,et al.  Time-series active search for quick retrieval of audio and video , 1999, 1999 IEEE International Conference on Acoustics, Speech, and Signal Processing. Proceedings. ICASSP99 (Cat. No.99CH36258).

[2]  Biing-Hwang Juang,et al.  Fundamentals of speech recognition , 1993, Prentice Hall signal processing series.

[3]  S. Furui,et al.  Cepstral analysis technique for automatic speaker verification , 1981 .

[4]  Kjell Lemström,et al.  SEMEX - An efficient Music Retrieval Prototype , 2000, ISMIR.

[5]  S. Boll,et al.  Suppression of acoustic noise in speech using spectral subtraction , 1979 .