| > Model Principle |
|
The underlying concepts of the proposed algorithms for perceptual measurement techniques are all quite similar. The common structure of these algorithms is depicted in the figure below. The process of human perception is modeled by employing a differential measurement technique which compares both, a reference signal (i.e. the "input" signal to a codec) and a test signal (i.e. the "output" signal of the codec). First, the algorithms process an ear model for the reference and the test signal, in order to calculate an estimate for the audible signal components. The result can be imagined as the "internal representation" inside the human auditory system. Comparing the internal representations of the reference, and the test signal leads to an estimate of the audible difference. To derive an overall quality figure, this information, which is a function of time, must be processed accordingly, like the human brain of a subject would do in a listening test. The respective part of processing within an algorithm is referred to as cognitive modeling. A total quality figure will be derived as the final result, which can be compared to a MOS ("Mean Opinion Score") resulting from a listening test.
|
|
|
The evaluation of the internal representation is often related to an estimate of the masked threshold. This estimate is based on data found in a number of psycho-acoustic experiments, such as those conducted by Zwicker [ZWIC67, ZWIC82]. Most of these experiments model certain isolated effects of the human auditory system. One way to design a perceptual measurement algorithm is to generalize these model data and apply them to complex audio signals. This was for example the approach outlined in the NMR Algorithm in 1987 [BRAN87, BRAN89, BRAN92, GILC96, HERR92a, HERR92b, KEYH93, KEYH96, KEYH98, SEIT89]. Similar approaches were used for PAQM, PSQM, PEAQ and PESQ [BEER95, BEER92, BEER94].
|
