Perceptual audio coding is used to achieve compact digital representations of high-fidelity audio signals for the purpose of efficient storage or transmission. Central objective in perceptual audio coding is the representation of the signal with a low number of bits while achieving high audio quality perception. Audio signals sampled at 44,1 kHz and 16 bit satisfy high quality demands-but to the disadvantage of high data rates when represented as PCM or as lossless compressed format (typically 706 kbit-1411 kbit/s for stereo signals). To attain high compression ratios, both irrelevancy and redundancy are removed from the signal (‘lossy compression’).

In this thesis, the development of a new audio coding format will be explained, which works as follows: Transformation into the frequency domain via Modified Discrete Cosine Transform (MDCT), decimation of the audio spectrum according to properties of the human ear, psychoacoustically optimized requantization using adaptive scalar quantization and vector quantization, followed by redundancy reduction. Target bitrates for this audio coder are 10 kbit/s to 14 kbit/s for mono and 12 kbit/s to 16 kbit/s for stereo signals, the stages are highly adapted to each other to attain high audio quality at these bitrates. It can be shown that due to the structure of the new audio coder the audio quality at these bitrates outperforms existing coders in subjective listening tests (example: mono@11,5 kbit/s: average score of 61% compared to existing formats in mono@16 kbit/s: 52% on MUSHRA- ratings according to the ‘Basic Audio Standard’ quality definition (ITU-R BS.1284) and 95% confidence interval).