Abstract: Session ITT-9

ITT-9.1	A realtime software MPEG transcoder using a novel motion vector reuse and a SIMD optimization techniques Yuzo Senda, Hidenobu Harasaki (NEC Corporation) A realtime software MPEG transcoder has been developed. A novel motion vector reuse and a SIMD optimization techniques are introduced to accelerate the transcoder without any quality degradation. Mean absolute error approximation criteria are employed in the reuse technique to refine scaled motion vectors. The developed transcoder on Pentium II 266MHz runs 2.5 times as fast as realtime, when scaling an MPEG-1 bitstream to half size.
ITT-9.2	Distributed Processing in the Home using a PC with a Wireless Speech Interface Debashis Chowdhury (Intel Corp., CEL), Ser J Chia (Intel Corp,' CEL) The Personal Computer is evolving from a purely personal device to one that can support multiple applications in different locations simultaneously. This paper describes how the connectivity and processing capability of the PC can be used in a distributed manner in the home to provide a variety of services like speech activated environmental command and control functions, digital video decoding, Internet telephony and entertainment control. As they are architected today, current PC's are challenged when trying to perform intensive signal processing tasks while managing several external connections (e.g. dial-up internet) and multiple internal connections (e.g. cordless phone interface) at the same time. We will describe some of these challenges, and what remains to be done to make the PC more capable in undertaking such a multifunctional challenge.
ITT-9.3	SPEECH RECOGNITION OVER THE INTERNET USING JAVA Zhemin Tu (Department of Applied Science, University of Arkansas at Little, Little Rock, AR 72204), Philipos C Loizou (Department of Applied Sciences, University of Arkansas at Little, Little Rock, AR 72204) A speech recognition system based on an Internet client -server model is presented in this paper. A Java applet records the voice at the client computer, sends the recorded speech file over the Internet, and the server computer recognizes the speech and displays the recognized text back to the user. Using this structure, an isolated digit recognition application was realized.
ITT-9.4	Implementation of a High-Quality Dolby* Digital Decoder Using MMX™ Technology James C Abel, Michael A Julier (Intel Corporation) Software decoding of Dolby Digital allows it to become a baseline capability on the PC, with greater flexibility than a hardware approach. Intel's MMX™ technology provides instructions that can significantly speed up the execution of the Dolby Digital decoder, freeing up the processor to perform other tasks such as video decoding and/or audio enhancement. Intel has worked closely with Dolby Laboratories to define an implementation of Dolby Digital based on MMX technology that has achieved Dolby's certification of quality.
ITT-9.5	Joint Source Channel Coding of Images over Frequency Selective Channels using DCT and Multicarrier BPAM Venceslav Kafedziski (Arizona State University) A novel approach to joint source and channel coding for frequency selective channels is presented. Multicarrier modulation is used to obtain an equivalent vector channel to the frequency selective channel and utilize the linear coding procedure of Lee and Petersen. The use of the block pulse amplitude transmission results in graceful degradation of the decoded signal for low channel SNR. The new procedure shows very good performance in the very low channel SNR region, for image transmission over frequency selective channels with deep nulls in the frequency response. Both encoder and decoder are computationally very inexpensive in terms of design and implementation, compared to the digital transmission with channel optimized vector quantization. Results for transmission of Gauss-Markov source and "Lena" image on several typical channels are presented.
ITT-9.6	Multi-rate speech coding for wireless and Internet applications John E Kleider (Motorola, Systems Solutions Group), Richard J Pattison (Motorola) Fixed-rate speech codecs are unable to provide synthesized speech with fixed delay when the channel capacity changes, and can not dedicate additional forward error correction bits for protection against noisy channels. We propose a multi-rate method for variable bandwidth applications, such as the Internet, and severely degraded wireless channels, such as mobile cellular. The technique uses a multi-rate version of the sinusoidal transform coder (MRSTC), operates at 9.6/4.8/2.4/1.2 kilobits/sec (kb/s), and is switchable "on-the-fly." The algorithm produces high quality speech, even when transitioning between rates. We compare two switching techniques, one method uses a "frame-deletion" (FD) technique, and a second method which utilizes "parameter-history" (PH) information. PH produces the best speech quality. FD is attractive because it requires no additional speech memory. Experimental results show greater than a 9 dB gain in receiver C/No operating range using the MRSTC over a fixed-rate system with STC operating at 9.6 kb/s.

< ITT-8

ITT-10 >

Last Update:  February 4, 1999         Ingo Höntsch