Events Calendar

Speech Coding and the BroadVoice Speech Codec

Speaker: Juin-Hwey (Raymond) Chen, Ph.D., Broadcom Corporation

Talk Title: Speech Coding and the BroadVoice Speech Codec

Abstract: This talk consists of two parts: Part 1 gives a brief introduction to speech coding in preparation for Part 2, which describes the BroadVoice® speech codec that was developed by Broadcom and was included in multiple standards for Voice over IP (VoIP) in cable telephony as standardized by CableLabs, SCTE (Society of Cable Telecommunications Engineers), ANSI, and ITU-T. Speech coding is the science and art of reducing the bit-rate needed to represent speech signals while maintaining a certain level of speech quality under given constraints on coding delay and complexity. Previous medium- to low-bit-rate speech coding standards typically have medium to high coding delay and complexity, and are usually encumbered by royalties. In contrast, the BroadVoice speech codec was designed from the outset to be low-delay, low-complexity, high-quality, and most importantly, free of third-party intellectual properties so it can be freely used by anyone without royalty payments. This talk gives a brief historical overview of the development of BroadVoice and also provides a high-level description of the BroadVoice speech coding algorithm. The BroadVoice family of codecs includes a 16 kb/s BroadVoice16™ (BV16) narrowband codec and a 32 kb/s BroadVoice32™ (BV32) wideband codec. Based on a novel two-stage noise feedback coding structure with vector quantization of excitation, BroadVoice achieves a low algorithmic buffering delay of merely 5 ms and codec complexity of only 12 and 17 MIPS for BV16 and BV32, respectively. The speech quality of BV16 is better than that of the ITU-T G.728, G.729, and 32 kb/s G.726 standards, and the speech quality of BV32 is better than that of the 64 kb/s G.722 standard. Audio demonstrations will be played in this talk. Both BV16 and BV32 are not only royalty-free but also open source with both floating-point and fixed-point C source code freely downloadable (www.broadcom.com/broadvoice).

Biography: Juin-Hwey (Raymond) Chen received his B.S.E.E. degree from National Taiwan University in 1980 and his Master and Ph.D. degrees in Electrical Engineering from University of California, Santa Barbara in 1983 and 1987, respectively. He joined Broadcom Corporation in 2000 and is now a Senior Technical Director there. Prior to joining Broadcom, Raymond spent 8 1/2 years at AT&T Bell Labs and AT&T Labs, three years at Voxware, Inc., and one year at Lucent Technologies. His research focus is on compression and quality enhancement of speech and audio signals. He has more than 50 technical publications and is an inventor of 57 issued United States patents and 35 pending U.S. Patents. He is the primary inventor of the ITU-T G.728 speech coding standard and the BV16 and BV32 speech codecs in the PacketCable, SCTE, ANSI, and ITU-T J.161 and J.361 standards. He is also well-known for inventing a postfiltering speech enhancement technique that is used in most international speech coding standards established since 1987. Throughout his career he has invented numerous speech codecs, many of which are widely used commercially. Raymond was elected a Fellow of the IEEE in 1995. He was also elected a Broadcom Fellow in 2006.

Host: Professor Shrikanth Narayanan

Location: Ronald Tutor Hall of Engineering (RTH) - 320

Audiences: Everyone Is Invited

Contact: Mary Francis