Afsaneh Asaei
Postdoctoral Researcher
Speech & Audio Processing
IDIAP Research Institute
Switzerland
Biography
I am a postdoctoral researcher at Idiap Research Institute, with supervision of Prof. Herve Bourlard. The subject of my research is parsimonious and deep learning for robust speech processing and recognition. The past fourteen years of my professional work have been devoted to various disciplines in signal processing and machine learning with a focus on deep learning, sparse component analysis, microphone arrays and statistical pattern recognition. Currently, my passion is parsimonious representation learning to achieve robustness in large scale applications. That is identifying an invariant space where machine listening paradigms can be realized for adverse (e.g. multiparty) conditions across many application domains.
Research Interest
My research interests lie in the broad area of signal processing, machine learning, statistics, acoustics, computational neuroscience, auditory and visual scene analysis for perception and cognition, sparse signal recovery and acquisition.
Publications
-
“Convexity in Source Separation: Models, geometry and algorithmsâ€, with M. B. McCoy, V. Cevher, Q. T. Dinh, L. Baldassarre, IEEE Signal Processing Magazine, Special Issue on Source Separation and Applications, 31 (3), 87-95, 2014.
-
“Structured Sparsity Models for Reverberant Speech Separationâ€, with M. Golbabaee, H. Bourlard, V. Cevher, IEEE/ACM Transactions on Speech and Audio Processing, , 22 (3), pp 620-633, 2014.
-
“Sparse Modeling of Neural Network Posterior Probabilities for Exemplar-based Speech Recognitionâ€, with P. Dighe, H. Bourlard, Speech Communication, vol 76, 230-244, 2016.