A decision of whether pronunciation is appropriate is based on the subjective judgment of a human supervisor by analyzing the acoustic features of speech. Commonly, it is made via a spectrogram. Loosely speaking, this method is based on the property that spectral changes that perceptually make the sounds differently should be associated with large distances and similarly keep the sound the same, perceptually should be associated with small distances. Spectral changes that perceptually lead to the sounds judged as being different are as follows:
Perceptual discrimination is helpful in relating a measurement of spectral difference to a perceived distortion, potentially making comparisons of speech patterns subjectively more meaningful. However, perceptual consistency itself is hard to obtain among human judges[18]. This concept can be applied not only to pronunciation evaluation, but also the feedback method in the L2 training utilizing the acoustic properties as described in section 2.2.4.
