Deep Scattering Spectrum with deep neural networks
Vijayaditya Peddinti, Tara N. Sainath, et al.
ICASSP 2014
Uniform quantizers are often modeled as additive uncorrelated noise sources. This paper explores the validity of the additive noise model in the environment of time-interleaved A/D converters. Cross-channel quantizer error correlation is an important discrepancy that arises for channel time delays in close proximity. It is demonstrated through simulation that negative error correlation occurs for different granularity quantizers in close proximity. Statistical analysis is presented to characterize error correlation between quantizers with different granularity. A technique exploiting this correlation often yields significant performance gains above the optimal additive noise model solution. © 2011 IEEE.
Vijayaditya Peddinti, Tara N. Sainath, et al.
ICASSP 2014
Shay Maymon, Etienne Marcheret, et al.
INTERSPEECH 2013
Shay Maymon, Pierre Dognin, et al.
INTERSPEECH 2013