Understanding the Differences Behind One Step and Two Steps Strategies in Speech Quality Evaluation

Strategies used in methods for objective speech quality evaluation can be classified into two main categories namely one step and two-step strategy. The one-step strategy aims at an algorithm or structure which transforms speech representation directly into a scalar, and the two-steps strategy often divides the evaluation procedure into two subsequent stages. Generally speaking, it constructs firstly objective measurements predefined for the disturbance within speech signal, and then accumulates the measurements to a single value which is similar to the opinion score. The two-step strategy has been developed ahead of one step strategy, leading to the study of the one step strategy. This paper describes the principles and essential elements in these strategies, compares the differences between them, and analyses the merits and drawbacks in these methods. Some criteria regarding to the method selection in speech quality evaluation which are based mainly on the research scenario are suggested at the end of the paper.