14th International Congress of Phonetic Sciences (ICPhS-14)

San Francisco, CA, USA
August 1-7, 1999


Perception of Noise Corrupted Speech: Practical Methods for Intelligibility and Quality Improvement.

Serguei Koval, Olga Ilina, Michael Khitrov, Vladimir Zaitsev

Speech Technology Center, Saint Petersburg, Russia

In some field applications experts have to decode speech of poor quality. The speech material to decode is often degraded quite seriously and can't be improved or changed to another, ‘clear’ copy. This report describes some principles of noisy speech recordings processing for improvement of message intelligibility and quality and some methods of speech enhancement for speech decoding.
   The discussed approaches have a good working record after many-years of practical use in STC and other institutions which perform noisy speech records decoding by orders of courts, prosecutors, law enforcement, emergency and accident investigation services, etc.
   Among main principles of distorted speech recordings preprocessing and noise cancellation we could note unmasking and normalization in time and frequency areas. Among main listening expert methods we could note selective listening, non-standard binaural listening accounting of short-term and long-term human ear adaptation to noisy speech.

Full Paper

Bibliographic reference.  Koval, Serguei / Ilina, Olga / Khitrov, Michael / Zaitsev, Vladimir (1999): "Perception of noise corrupted speech: practical methods for intelligibility and quality improvement.", In ICPhS-14, 803-806.