Quality Assessment

Subjective

Proper subjective assessment of audio quality is usually done using a test suite of reference music and speech recordings [15]. These tests require strictly controlled listening conditions [16], and a double blind listening procedure (for example, according to ITU-R Recommendation BS.1116-1 for evaluation of high quality systems exhibiting small impairments, a double-blind triple-stimulus test with hidden reference is used [17]). For reliable comparison of different systems, the assessment should be done by several listeners, and the results should be statistically analyzed. The score is usually scaled using six-grade Impairment Scale (according to ITU-R Recommendation BS.562-3 and the new ITU-R BS.1284 [18]) or five-interval Continuous Quality Scale (according to ITU-R Document 10-11Q/3 [19]). To learn more about subjective quality assessment, see [16, 20].

Objective

The amount of nonlinear distortions and external noise and artifacts introduced by given system can be measured [21,22] (and compared to another system) by processing a reference sinusoidal signal and using two simple quality measures:
total harmonic distortion: ,   and signal to noise ratio: ,
where P denotes the total power of the output distorted signal, and P_n denotes the power of its n-th harmonic (n=1 correspond to the fundamental which is equivalent to the input sinusoid).