University of Louisville
Speech Recognition in Time-Forward and Time-Reverse Conditions
Institution
University of Louisville
Faculty Advisor/ Mentor
Pavel Zahorik
Abstract
Daily we are exposed to speech in varying settings and room conditions. All rooms that we listen in are not ideal for speech. Previous studies suggest that several characteristics contribute to listening within different environments, including familiarity with the listening environment. In the current study familiarity was examined through by comparing performance on a speech recognition task in two listening conditions; a normal time-forward condition and unfamiliar time-reverse condition. Using virtual auditory space speech was presented through headphones giving the perception of a realistic room environment that preserves reverberant conditions but allowed manipulation of stimulus characteristics. Current analyses indicate a decrease in word recognition for room reverse conditions. Results of percent correct word recognition yielded approximately 80% for room-forward, and less than 20% for room-reverse. This dramatic decrease may perhaps be attributed to a lack of familiarity with the environment or degradation of onset information. The current research may contribute to development of optimal listening environments and assisted listening technology.
Speech Recognition in Time-Forward and Time-Reverse Conditions
Daily we are exposed to speech in varying settings and room conditions. All rooms that we listen in are not ideal for speech. Previous studies suggest that several characteristics contribute to listening within different environments, including familiarity with the listening environment. In the current study familiarity was examined through by comparing performance on a speech recognition task in two listening conditions; a normal time-forward condition and unfamiliar time-reverse condition. Using virtual auditory space speech was presented through headphones giving the perception of a realistic room environment that preserves reverberant conditions but allowed manipulation of stimulus characteristics. Current analyses indicate a decrease in word recognition for room reverse conditions. Results of percent correct word recognition yielded approximately 80% for room-forward, and less than 20% for room-reverse. This dramatic decrease may perhaps be attributed to a lack of familiarity with the environment or degradation of onset information. The current research may contribute to development of optimal listening environments and assisted listening technology.