v7.0, 20040922, me

Thursday, October 28, 1:00 pm – 4:00 pm
Session D LOSSY AND LOSSLESS AUDIO CODING, Part 2

Chair: James Johnston, Microsoft Corporation, Redmond, WA, USA

1:00 pm
D-1 Introduction to Dolby Digital Plus, an Enhancement to the Dolby Digital Coding System—Louis D. Fielder, Robert L. Anderson, Brett G. Crockett, Grant A. Davidson, Mark F. Davis, Stephen C. Turner, Mark S. Vinton, Phillip A. Williams, Dolby Laboratories, San Francisco, CA, USA
An extension to the existing Dolby Digital (AC-3) multichannel audio coding standard is described and its new capabilities explored. This coding system is designed with extensive compatibility with the existing standard by retaining most of its metadata and data-framing structure to preserve and extend functionality in existing applications. New features include simultaneous support for multiple program streams, carriage of multichannel signals beyond 5.1 channels, and fine-grained control and support for data rates up to 6.144 Mbps. New coding “tools” including spectral extension, enhanced channel coupling, transient pre-noise processing, and improved filterbank/quantization enhance the performance of earlier AC-3 technology.
Convention Paper 6196

1:30 pm
D-2 Ultra Low Delay Audio Coding with Constant Bit Rate— Ulrich Krämer, Gerald Schuller, Stefan Wabnik, Juliane Klier, Jens Hirschfeld, Fraunhofer Institute for Digital Media Technology, Ilmenau, Germany
The Ultra Low Delay (ULD) codec developed at the Fraunhofer IDMT is based on a versatile perceptual audio coding method that achieves very low encoding/decoding delay and is nevertheless capable of high compression ratios. Utilizing a perceptual model for irrelevance reduction, the ULD codec is in principle a variable bit rate codec. To achieve coding with constant bit rate, the use of bit reservoir techniques would result in additional coding delay. This paper presents a rate loop, which ensures constant bit rate coding without increasing coding delay. It is shown that this technique does not decrease the decoded audio quality significantly.
Convention Paper 6197

2:00 pm
D-3 An Analysis of Tandem Error During Audio Transcoding—Jun Wei Lee, National University of Singapore, Singapore; Aweke Lemma, Royal Philips Electronics, Eindhoven, The Netherlands; Michiel van der Veen, Royal Philips Electronics, Eindhoven, The Netherlands
In a repeated quantization scenario, apart from the nominal quantization error, an additional tandem error is introduced. The amount of the excess tandem error depends on the characteristics of the quantizers used. In this work, the effect of tandem error and its dependence on the underlying quantizer characteristics are analyzed. A prime example where tandem error leads to increased noise is in audio transcoding. The behavior of tandem-noise for typical audio coding methods such as MPEG 1 Layer 2 is investigated. A method of reducing the tandem errors is proposed. This method involves guiding the quantization process of the first quantizer, assuming prior knowledge of the second quantizer. Results show that the method is able to reduce the amount of tandem error in the repeated quantization scenario.
Convention Paper 6198

2:30 pm
D-4 aacPlus, Only a Low-Bit-Rate Codec?—Andreas Ehret, Coding Technologies, Nürnberg, Germany; Kristofer Kjörling, Coding Technologies, Stockholm, Sweden; Jonas Rödén, Coding Technologies, Stockholm, Sweden; Heiko Purnhagen, Holger Hörich, Coding Technologies, Nürnberg, Germany
aacPlus, the combination of the well known MPEG AAC and the Spectral Band Replication tool SBR has been introduced as a highly efficient low bit-rate audio codec, representing today’s state-of-the-art by providing full bandwidth, near CD audio quality at 48 kbit/s stereo. It is thus suited for applications that demand the highest compression ratios. This paper discusses benefits when using aacPlus at moderate compression ratios in the range of 80 to 128 kbit/s stereo, where so far AAC was the codec of choice. The technological approach for applying SBR in such a scenario is described and subjective evaluations of the presented solution as well as an overview on system and implementation aspects are given.
Convention Paper 6199

3:00 pm
D-5 Efficient Bit Reservoir Design for MP3 and AAC—Chi-Min Liu, National Chiao-Tung University, Hsin-Chu, Taiwan; Li-Wei Chen, National Chiao-Tung University, Hsin-Chu, Taiwan; Ming-Ton Su, National Chiao-Tung University, Hsin-Chu, Taiwan; Wen-Chieh Lee, National Chiao-Tung University, Hsin-Chu, Taiwan; Chung-Han Yang, National Chiao-Tung University, Hsin-Chu, Taiwan; You-Hua Hsiao, National Chiao-Tung University, Hsin-Chu, Taiwan; Zheng-Wen Li, InterVideo Digital Technology (Shanghai) Co., Ltd., ShangHai, China; Chu-Ting Chien, InterVideo Digital Technology, Taipei, Taiwan
Bit reservoir controlling the bits budget among music frames has been the kernel module to have good bit-quality tradeoff in current audio encoders like MP3 and AAC. The approaches of bit reservoirs can be investigated from demand-driven approach and budget-driven one. Demand-driven approach determines the required bits according to the audio contents while budget-driven one allocates bits according to the bit budgets accumulated in the bit reservoir. Existing bit reservoirs follow basically the budget-driven approach. This paper presents an efficient bit reservoir design with concerns from both demand and budget. The bit reservoir includes a demand estimator to adaptively predict the bits required for each frame. Also, the bit reservoir has a budget regulator to control the bits used according to the codec protocol and the preferred scenario. The new bit reservoir method is included in MP3 and AAC to verify the efficiency through extensive objective and subjective tests.
Convention Paper 6200

3:30 pm
D-6 Design of MPEG-4 AAC Encoder— Chi-Min Liu, Wen-Chieh Lee, Chung-Han Yang, Kang-Yan Peng, Ting Chiou, Tzu-Wen Chang, Yu-Hua Hsiao, Hen-Wen Hue, National Chiao-Tung University, Hsin-Chu, Taiwan; Chu-Ting Chien, NangKang Software Park, Taipei, Taiwan
The state-of-art natural audio coder, MPEG-4 AAC, has provided extensive coding modules for achieving high coding efficiency. The modules, which include filter bank, window switch, psychoacoustic model, bit allocation, bit reservoir, lossless coding, temporal noise shaping, and middle/side coding, span a large design dimension and create challenges for audio coding technology. In this paper the design of these modules, named a NCTUAAC encoder is presented to provide adequate audio quality with low computation complexity.
Convention Paper 6201