TY - CHAP
T1 - Source separation, localization, and comprehension in humans, machines, and human-machine systems
AU - Durlach, Nat
AU - Colburn, Steve
AU - Kidd, Gerald
AU - Mason, Chris
AU - Shinn-Cunningham, Barbara
AU - Arborgast, Tanya
AU - Devore, Sasha
AU - Gallun, Erick
AU - Zurek, Pat
AU - Desloge, Jay
PY - 2005
Y1 - 2005
N2 - At the most general level, the task being addressed in this book is that of processing the sound field in realistic acoustic environments (containing both multiple sources and reverberation) in a manner that facilitates goal-oriented behavior in these environments. In terms of classical terminology (most frequently used when human processing rather than machine processing is considered), the task is concerned with "Auditory Scene Analysis" and with factors underlying the "Cocktail Party Effect". Obviously included as sub elements of the specified task (beyond mere detection) are the tasks of (a) separating and localizing the acoustic sources, (b) distinguishing between characteristics of the received signals that are associated with the transmitted signals and those that are associated with the signal transformations (the filtering) imposed by the acoustic environment in which the sources and sensors are located, and (c) comprehending the sources (e.g., understanding the speech and determining the identity of all the talkers in the event that the sources consist of a number of people talking simultaneously). Although some of the relevant research is sufficiently general to apply to sources other than speech, the focus in this book is on speech communication. However, the types of systems to be considered with respect to the problem of speech reception in complex environments are very broad [including humans, machines (robots), and combinations of humans and machines (humanmachine systems)]. Such broad coverage is necessitated both by the multiple goals of the book (understanding humans and designing better machines) and by the belief that the productivity of research in this area can be greatly enhanced by the use of multidisciplinary teams attacking arrays of related problems in a common framework. Overall, it is anticipated that the kinds of research discussed in this book will advance both (1) our understanding of how humans solve the problems in question using their natural biological systems and (2) our ability to design improved machines or human-machine systems for solving these problems. Past research has clearly demonstrated that understanding human processing can benefit design of artificial processing and that knowledge of machine processing techniques can play an important role in understanding human processing. It is also clear that judicious combinations of human and machine processing can lead to systems that are superior to either type by itself, and that design of optimum humanmachine systems requires improved knowledge of the advantages and disadvantages of both types of systems. This chapter is divided into two parts. The first part contains some general comments about the tasks and the systems. The second focuses on two important highly relevant, currently active, research areas in human auditory perception: (A) reverberation and (B) informational masking.
AB - At the most general level, the task being addressed in this book is that of processing the sound field in realistic acoustic environments (containing both multiple sources and reverberation) in a manner that facilitates goal-oriented behavior in these environments. In terms of classical terminology (most frequently used when human processing rather than machine processing is considered), the task is concerned with "Auditory Scene Analysis" and with factors underlying the "Cocktail Party Effect". Obviously included as sub elements of the specified task (beyond mere detection) are the tasks of (a) separating and localizing the acoustic sources, (b) distinguishing between characteristics of the received signals that are associated with the transmitted signals and those that are associated with the signal transformations (the filtering) imposed by the acoustic environment in which the sources and sensors are located, and (c) comprehending the sources (e.g., understanding the speech and determining the identity of all the talkers in the event that the sources consist of a number of people talking simultaneously). Although some of the relevant research is sufficiently general to apply to sources other than speech, the focus in this book is on speech communication. However, the types of systems to be considered with respect to the problem of speech reception in complex environments are very broad [including humans, machines (robots), and combinations of humans and machines (humanmachine systems)]. Such broad coverage is necessitated both by the multiple goals of the book (understanding humans and designing better machines) and by the belief that the productivity of research in this area can be greatly enhanced by the use of multidisciplinary teams attacking arrays of related problems in a common framework. Overall, it is anticipated that the kinds of research discussed in this book will advance both (1) our understanding of how humans solve the problems in question using their natural biological systems and (2) our ability to design improved machines or human-machine systems for solving these problems. Past research has clearly demonstrated that understanding human processing can benefit design of artificial processing and that knowledge of machine processing techniques can play an important role in understanding human processing. It is also clear that judicious combinations of human and machine processing can lead to systems that are superior to either type by itself, and that design of optimum humanmachine systems requires improved knowledge of the advantages and disadvantages of both types of systems. This chapter is divided into two parts. The first part contains some general comments about the tasks and the systems. The second focuses on two important highly relevant, currently active, research areas in human auditory perception: (A) reverberation and (B) informational masking.
UR - http://www.scopus.com/inward/record.url?scp=84875873765&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=84875873765&partnerID=8YFLogxK
U2 - 10.1007/0-387-22794-6_15
DO - 10.1007/0-387-22794-6_15
M3 - Chapter
AN - SCOPUS:84875873765
SN - 1402080018
SN - 9781402080012
SP - 221
EP - 243
BT - Speech Separation by Humans and Machines
PB - Springer US
ER -