Speaker Bios
James Johnston
By
Aug 1, 2008, 13:22

JJ is currently employed at Neural Audio in Kirkland, Wa, where he is working on a variety of preprocessing and postprocessing algorithms for audio capture, analysis, control, and presentation.

Prior to his position at Neural Audio, he worked for 5 years at Microsoft Corporation in the "Codecs", "Core Media Processing" and finally the video services groups as Audio Architect.

He is retired from AT&T Labs - Research, headquartered at Florham Park, NJ, Speech Processing Software and Technology Research Department. Before that, he was employed by AT&T Bell Laboratories, in the Acoustics Research Department under Dr. J. L. Flanagan, and in the Signal Processing Research Department.

He started his career working on using analog signal processing to do speech coding (APCM, ADPCM, SBC) for testing of algorithms, sampling rates, and quantizer resolutions. His first IEEE paper detailed the hardware construction of an ADPCM implementation using analog multipliers and integrators to provide both step-size and predictor "calculation", in a form that allowed sampling rate and quantizer resolution changes.

Since then, he has worked in analog signal processing, speech coding, voice privacy, quadrature mirror filter design, and perceptual coding of both audio and images. During this work on perceptual audio coding, he has been the primary investigator of the early PXFM audio coder which was reported on at the ASSP Digital Audio Meeting in Mohonk, NY in 1986 and a co-inventor and standards proponent of the ASPEC algorithm, the quality leader in the MPEG-1 audio competition.

During this time, he also did an investigation of coding of still-frame images using a forward-driven perceptual model with Dr. R. J. Safranek, also of AT&T Bell Laboratories. This image coder, called PIC (for Perceptual Image Coder), used very simple techniques to provide state of the art still-image compression.

Perceptual coding, as compared to source coding, pays primary attention to the intended reciever or destination of the signal, rather than the source or source model of the signal. This destination is usually the human being. By eliminating imperceptable information in the signal, compression rates that are substantially better than the usual source coding rates are achieved, with substantially more mean squared error, but better percieved quality than the source coder.

Until 2002, he was the primary researcher and inventor of AT&T's contributions to the MPEG-2 AAC audio coding algorithm. He also represented AT&T in the ANSI accredited group X3L3.1, and X3L3.1 in the ISO-MPEG-AUDIO (AAC) arena in support of the AAC algorithm.