MPEG Audio Coding

After the advent of compact disk the end user has started to expect high quality output from any digital audio application. The increasing demand of audio transmission and exchange over the internet and the growing trend of storage/playback of high quality audio on digital computer systems persistently call for developing new and improved audio coding techniques which can meet the contradicting demands of high quality and low bit rate both at the same time. Perceptual audio coding which employs the process of irrelevancy reduction through exploitation of human auditory perception has a central place in modern audio coding scenario. In this thesis several new techniques are developed within the paradigm of perceptual audio coding to enhance the coding efficiency and produce better output quality at reduced bit rates. The performance of developed techniques is evaluated by employing both subjective and objective measures recommended by the ITU-R. Moreover a coding scheme is developed which integrates the advantages of separately available sophisticated speech and audio coding techniques within a single algorithm. This codec is intended to cater for the needs of present multimedia scenario which contains mixed speech and audio signals thus requiring a single codec to handle all signal types for optimal coding efficiency.

Efficient Audio Coding Employing Novel Techniques a Codec Selector Algorithm

A novel technique of Phase Randomization (PR) for bit rate reduction in perceptual coding of high quality audio is presented. Instead of actual phase information, the decoder utilizes nevertheless?carefully?formulated `random' phase information for signal reconstruction with the help of a code sent by the encoder in every frame of data. The coding scheme is tested on the Digital Audio Broadcasting (DAB) Standard, for which broadcasters often desire an increase in the number of transmission channels for densely populated areas. The codec achieves the aim of providing good quality audio at a reduced bitrate assessed with the help of ITU-R BS. 1116 methodology for the subjective assessment of audio quality.

Phase Randomization-A new concept in Perceptual Audio Coding

In this paper we present an efficient high quality perceptual audio coding scheme with a novel dynamically switched filter bank design and an associated bit allocation strategy. An additional processing block is added prior to the decomposition subband filter bank, where a decision for appropriate time/frequency resolution selection is taken based on the nature of the audio signal. A block-by-block analysis is then carried out to identify the prominent spectral peaks (PSPs), with a view to efficiently utilize their masking potential in bit rate reduction. A particular consideration is given to critical band boundaries while formulating uniform subbands, to prevent the injected coding noise becoming perceptible. A compact representation is achieved by selection of either a critical, uniform or non-uniform non-critical subband configuration depending on minimization of Perceptual Entropy (PE) estimates. An optimum bit allocation algorithm with a rate-distortion loop allots lesser bits to subbands with PSPs as compared to the actual bit demand. Test results based on ITU-R recommendations show that our coding scheme performs considerably better as compared to MPEG-1 Layer-3 (the de-facto standard for audio interchange on the Internet) for a majority of signal types. The paper concludes with a discussion of future research implications of the work.

http://www.wseas.us/e-library/conferences/2005corfu/c1/papers/498-409.pdf

Muhammad Tayyab Ali

Papers

Phase Randomization-A new concept in Perceptual Audio Coding

Efficient signal adaptive perceptual audio coding