[...]... feature in a speech stream—envelope fluctuations or formant trajectories—can be masked by random sequences of the same feature in another stream The chapter by Xiong and Huang steps out of the acoustic domain and shows how visual and auditory information by talkers speaking simultaneously interact and how this interaction can be successfully used by humans and machines to separate multiple speech sources... estimating various parameters of speech Stern is taking yet another different route: applying auditory models to recover speech, both by separation and by recognition—a hybrid bottom-up and top-down approach Irino, Patterson, and Kawahara show how speech separation can be achieved by using a combination of two sophisticated signal processing methods: the auditory image method (AIM) and STRAIGHT, a pitch-synchronous... models, and assumptions To a casual observer it seems that, despite commendable efforts and achievements by many researchers, it is not clear where the field is going At the same time, despite an accelerating increase in investigations by neuroscientists that have led to characterizing and mapping more and more of the auditory and cortical processes responsible for speech separation by man, our understanding... method is further described by Kawahara and Irino showing how STRAIGHT, a method consistent with auditory processing, can also lead to the separation of speech signals even by itself Auditory models are the explicitly stated base of CASA, the focus of several successive chapters The one by Wang and Hu suggests that separation of speech from a noisy background can be achieved by applying an estimated optimal... Rather, it is a matrix of data, background, and ideas that cuts across fields and approaches, with the unifying theme of separation and interpretation of speech corrupted by a complex acoustic environment Over and above presenting facts and accomplishments in this field, the landscape that the book wishes to paint also highlights what is still missing, unknown, and un-invented It is our hope that the book... First, we want to emphasize that at present the field of speech separation by human and machines is complex and replete with ill-defined and un-agreed-upon concepts Second, despite this, we want to express our hope that significant accomplishments in the field of speech separation may be come to light in the not-so-distant future, propelled by a dialog between proponents of different approaches, The... Ellis addresses one of the ubiquitous problems of speech recovery by machines: performance evaluation The final chapter by Cooke takes a broad view of speech separation Looking through the lens of someone committed to uncovering the mysteries of ASA and CASA, he proposes that time-frequency information in badly degraded portions of speech may be recovered by glimpsing at those regions where this information... relevant to speech understanding in interference It was therefore appropriate that Divenyi assembled members of a number of disciplines working on the problem of the separation of concurrent sounds: experimental psychologists studying how ASA was done by people, both for speech and non -speech sounds, neuroscientists interested in how the brain deals with sounds, as well as computer scientists and engineers... mind before and during the preparation of this book Lastly and most of all, however, I want to express my gratitude to Mary Harper, whose enthusiastic support of multidisciplinary research on speech separation was responsible for the workshop and has created the impetus for the book Pierre Divenyi Bell, A.J and Sejnowski, T.J., 1995, An information maximisation approach to blind separation and blind... engineers develop- xvi Speech Separation ing computer systems to solve the problem This book is a fascinating collection of their views and ideas on the problem of speech separation My personal interest in these chapters is that they bring to forefront an argument of special import to me as a cognitive psychologist This argument, made by CASA researchers, is that since people can do sound separation quite . +08'00' SPEECH SEPARATION BY HUMANS AND MACHINES This page intentionally left blank SPEECH SEPARATION BY HUMANS AND MACHINES Edited by Pierre Divenyi East Bay Institute for Research and Education KLUWER. Smaragdis 83 Automatic Speech Processing by Inference in Generative Models Sam T. Roweis 97 viii Speech Separation Signal Separation Motivated by Human Auditory Perception: Applica- tions to Automatic Speech. For Robust Speech Recognition in Noisy and Reverberant Conditions 213 Guy J. Brown and Kalle J. Palomäki Source Separation, Localization, and Comprehension in Humans, Ma- chines, and Human-machine