Unsupervised and Neural Hybrid Techniques for Audio Signal Classification
Andrs Ortiz, Javier Lorenzo-Navarro, Ana M., Isabel Barbancho · InTech eBooks · 2012
Audio signal analysis and classification have arisen as an important research topic that has been developed by many authors in different areas over the time.A main development context has been speech recognition Holmes & Huckvale (1994); Juang & Rabiner (2005); Kimura (1999).This specific topic is, in fact, an important source of references to applications of artificial intelligence techniques Juang & Rabiner (2005); Prasad & Prasanna (2008).Many classical problems encountered in this research field have been addressed from this perspective by many authors Farahani & Ahadi (2005); Minematsu et al. (2006).However, in this same context, a different point of view can be adopted to deal with the analysis and classification of music signals.The tasks in this framework are varied, including the detection of pitch, tempo or rhythm Thornburg et al. (2007), but also other tasks like the identification of musical instruments Müller et al. (2011) or musical genre recognition Tzanetakis & Cook (2002) can be considered.Also the classification of audio samples as music or speech has been thoroughly considered in the literature Panagiotakis & Tziritas (2005), Tardón et al. (2010).In this specific context of analysis of musical signals, we will expose some ideas regarding signal classification and their application to a real task. Models and applicationsAudio signal classification involves the extraction of a number of descriptive features from the sound and the proper utilization of them as input for a classifier.Artificial Intelligence (AI) techniques provide a way to deal with signal processing and pattern classification tasks from a different point of view of classical techniques.AI techniques play an important role in the signal processing and classification context as they have been widely used by a number of authors in the literature for very different tasks Haykin (1999); Kohonen et al.