Standardization activities within MPEG-4 have been started which aim at the developement of a coding standard for transmitting audio signals at bit rates below 16 kbit/s.
Currently available general purpose audio codecs operate at bit rates above 16 kbit/s. On the other hand speech codecs operating in the desired range of bit rates are not suited for encoding audio signals.
A coding scheme is to be developed wich allows the transmission of audio signals with a bandwidth of 4 ... 8 kHz at bit rates between 6 and 16 kbit/s.
The concept of an object based analysis/synthesis scheme is applied.
In the encoder the input signal is first decomposed into audio objects which can be described by appropriate source models and represented by corresponding sets of model parameters. For each object the model parameters are estimated, coded, and transmitted. Additionally a signal is synthesized from the derived parameters which assists the decomposition of the residual. Quantization and coding of the model parameters is controlled by a psychoacoustic model which exploits the input signal dependent sensitivity of the human ear to reconstruction errors.
In the decoder for each audio object a signal is synthesized from the decoded model parameters. The decoder output is the sum of all object signals.
This coding scheme was adopted as parametric audio coder in the MPEG-4 Audio Standard, where it is referred to as HILN ("Harmonic and Individual Lines plus Noise").
Further details about this coding scheme are available from numerous publications. For general information, see the Parametric Audio Coding Bibliography.
