The algorithm performs a psychoacoustical analysis of frequency components of the signal and produces the SMR value for each subband and every channel of the audio signal. The SMR value is obtained from the difference between the highest signal level and the lowest level of the masking threshold in each subband. Computational accuracy depends on the chosen layer.
The SMR value is used to determine the quantity of bits necessary to quantize the subband samples, exploiting the fact that from a perceptive standpoint it does not make sense to have a SNR value higher than MNR.
For psychoacoustical model 1, computation of SMR proceeds as follows:
| 1 - | Computation of the FFT to represent PCM samples in the frequency domain |
| 2 - | Computation of the Sound Pressure Level (SPL) in dB for each subband |
| 3 - | Comparison of the static threshold for no signal (precomputed in a table) |
| 4 - | Identification of tonal (similar to sine waves) and non-tonal (similar to noise) components |
| 5 - | Selection of relevant masking tones |
| 6 - | Computation of the individual masking threshold for every spectral component |
| 7 - | Computation of the global threshold |
| 8 - | Computation of the minimum of the masking threshold for each subband |
| 9 - | Computation of the SMR in each subband |
All these steps will be examined in detail for a sampling frequency of 48 kHz (sampling frequency influences only the contents of some tables of coefficients, but does not have anything to do with the computational process) and for layer II.
[ Index | Main Paragraph | Next Paragraph ]