Abstract
In this chapter, we delve into the problem of relative transfer function (RTF) identification. First, we focus on identification algorithms that exploit specific properties of the input data. In particular, we exploit the non-stationarity of speech signals and the existence of segments where speech is absent in arbitrary utterances. Second, we explore approaches that aim at better modeling the signals and systems. We describe a common approach to represent a linear convolution in the short-time Fourier transform (STFT) domain as a multiplicative transfer function (MTF). Then, we present a new modeling approach for a linear convolution in the STFT domain as a convolution transfer function (CTF). The new approach is associated with larger model complexity and enables better representation of the signals and systems in the STFT domain. Then, we employ RTF identification algorithms based on the new model, and demonstrate improved results.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Preview
Unable to display preview. Download preview PDF.
Author information
Authors and Affiliations
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2010 Springer Berlin Heidelberg
About this chapter
Cite this chapter
Talmon, R., Cohen, I., Gannot, S. (2010). Identification of the Relative Transfer Function between Sensors in the Short-Time Fourier Transform Domain. In: Cohen, I., Benesty, J., Gannot, S. (eds) Speech Processing in Modern Communication. Springer Topics in Signal Processing, vol 3. Springer, Berlin, Heidelberg. https://doi.org/10.1007/978-3-642-11130-3_2
Download citation
DOI: https://doi.org/10.1007/978-3-642-11130-3_2
Publisher Name: Springer, Berlin, Heidelberg
Print ISBN: 978-3-642-11129-7
Online ISBN: 978-3-642-11130-3
eBook Packages: EngineeringEngineering (R0)