Grant Allen Davidson

High-quality audio transform coding at 128 kbits/s

An approach to wideband digital audio compression of CD-quality signals at data rates of 128 kb/s channel and below is presented. A form of adaptive transform coding, this technique features a nonuniform frequency division and coding scheme to exploit known characteristics of human perception. The algorithm has low computational complexity and can be adapted for use at other bit rates. A windowed overlap-add process is used with the forward/inverse transforms, which have been efficiently implemented using FFTs. Transform coefficients are converted into a subband block-companded format consisting of exponent words and associated mantissas, which are then coded with an adaptive quantizer. A real-time, single-chip programmable digital signal processing (DSP) implementation encodes 480-kHz-sampled stereo audio signals at a variety of bit rates. At 128 kb/s, the coders subjective performance is appropriate for highest-quality 15-kHz professional audio applications.<<ETX>>

ATSC Video and Audio Coding

In recent decades, digital video and audio coding technologies have helped revolutionize the ways we create, deliver, and consume audiovisual content. This is exemplified by digital television (DTV), which is emerging as a captivating new program and data broadcasting service. This paper provides an overview of the video and audio coding subsystems of the Advanced Television Systems Committee (ATSC) DTV standard. We first review the motivation for data compression in digital broadcasting. The MPEG-2 video and AC-3 audio compression algorithms are described, with emphasis on basic concepts, system features, and coding performance. Next-generation video and audio codecs currently under consideration for advanced services are also presented.

A low cost adaptive transform decoder implementation for high-quality audio

Low-cost, 16-b fixed-point digital signal processing (DSP) chips have traditionally been used in real-time wideband audio implementations due to limited arithmetic precision, which can lead to audible roundoff errors. The authors describe a real-time AC-2 stereo digital audio decoder implementation based on one 16-b TMS320C5x DSP. This is achieved by modifying a conventional inverse fast Fourier transform (FFT) computation, using a form of mixed-precision arithmetic, and exploiting the short instruction cycle time of the DSP. Compared with a 16-b single-precision implementation, a moderate increase in required DSP cycle time is incurred. The results indicate that the dynamic range of the 16-b DSP decoder is currently within 1 to 3dB of that obtained by current high-quality 16-b A/D and D/A converters. A further refinement will produce a dynamic range figure which meets or exceeds that obtained by higher-precision fixed-point ALUs.<<ETX>>

Time versus Frequency Resolution in a Low-Rate, High Quality Audio Transform Coder

An adaptive block size transform coder for high quality music has been developed. The adaptability of the input size of the transform combined with the properties of the transform as developed in the Dolby AC-2 technology allows one to exploit both maximum time and frequency resolution while keeping the bit rate as low as 128 kb/s per channel. The low complexity of the system permits a real-time implementation of encoder or decoder using one general purpose, programmable DSP chip per channel pair.

Multidimensional Optimization of MPEG-4 AAC Encoding

The subjective quality achieved by most audio codecs, including MPEG-4 AAC, depends strongly on the algorithms used for encoder parameter selection. As a practical measure in conventional encoders, the overall encoding procedure is usually divided into a sequence of smaller problems that are solved heuristically. In this paper, we formulate the MPEG-4 AAC encoding problem as a multidimensional optimization procedure and present simulation results indicating performance gains relative to conventional approaches

Transform-domain decorrelation in Dolby Digital Plus

The Dolby Digital Plus (DDPlus) multichannel audio codec employs channel coupling for parsimonious transmission of high frequency components of the signal, wherein the transform coefficients of discrete channels beyond a coupling-begin frequency are transmitted as a mono-downmix. The decoder reconstructs individual channels by appropriate panning with frequency-banded gains. While channel coupling is a valuable parametric coding tool for low bit-rate encoding of multichannel audio, the resulting high inter-channel coherence can affect audio post-processors that utilize downmixing, such as headphone virtualizers. To mitigate these effects, a new low-complexity spatial audio coding tool is proposed, whose primary objective is to restore phase diversity in the decoders output. In this new approach, a decorrelation signal is synthesized directly from the decoded real-valued and critically-sampled transform (MDCT) coefficients, avoiding the expense of computing the imaginary counterpart (MDST) or signal transformation to a different domain (such as QMF). The decorrelation signal is then adaptively mixed with the dry signal and inverse-transformed as in a conventional decoder. The degree of mixing depends on spatial parameters that are either sent in the bitstream or estimated in the decoder from the discrete (not coupled) low-frequency spectral coefficients. Listening tests demonstrate the significant performance benefits obtained in either mode of operation of the proposed tool.

Quantization in perceptual audio coders with correction for synthesis filter bank noise spreading

Perceptual transform and subband audio coding systems using analysis/synthesis filter banks introduce noise by quantizing subband signals or transform coefficients. Historically, systems have been designed under the assumption that the quantization noise power spectra at the synthesis filter bank input and output are equal. In general, the power spectra will have different shapes since the synthesis convolution leads to quantization noise spreading. In this paper, a theoretical framework for deriving an optimum noise allocation that accounts for spreading by the synthesis filter bank is described. In concept, the problem of finding an optimal noise allocation can be expressed as a constrained optimization problem. A simplified algorithm derived from the theoretical framework is discussed that was found to improve subjective audio quality while using only modest computational resources.