Paper Info
Title
Binaural speech separation using recurrent timing neural networks for joint F0-localisation estimation
Abstract
A speech separation system is described in which sources are represented in a joint interaural time difference-fundamental fre- quency (ITD-F0) cue space. Traditionally, recurrent timing neural net- works (RTNNs) have been used only to extract periodicity information; in this study, this type of network is extended in two ways. Firstly, a coincidence detector layer is introduced, each node of which is tuned to a particular ITD; secondly, the RTNN is extended to become two- dimensional to allow periodicity analysis to be performed at each best- ITD. Thus, one axis of the RTNN represents F0 and the other ITD allow- ing sources to be segregated on the basis of their separation in ITD-F0 space. Source segregation is performed within individual frequency chan- nels without recourse to across-channel estimates of F0 or ITD that are commonly used in auditory scene analysis approaches. The system is evaluated on spatialised speech signals using energy-based metrics and automatic speech recognition.
Year
DOI
Venue
2007
10.1007/978-3-540-78155-4_24
Lecture Notes in Computer Science
Keywords
Field
DocType
individual frequency channel,periodicity information,itd-f0 space,automatic speech recognition,particular itd,cue space,auditory scene analysis approach,spatialised speech,speech separation system,binaural speech separation,periodicity analysis,f0-localisation estimation,recurrent timing neural network
Auditory scene analysis,Pattern recognition,Computer science,Pitch period,Communication channel,Speech recognition,Artificial intelligence,Coincidence detection in neurobiology,Artificial neural network,Binaural recording
Conference
Volume
ISSN
ISBN
4892
0302-9743
3-540-78154-4
Citations 
PageRank 
References 
3
0.41
9
Authors
2
Name
Order
Citations
PageRank
Stuart N. Wrigley118120.56
Guy J. Brown276097.54