Research on the Objective and Subjective Evaluation Methods of Speech Quality Driven by Feature Fusion

Abstract

Aiming at the problem of inaccurate speech quality evaluation caused by the previous subjective evaluation-dominated and insufficient coverage of objective evaluation indicators in military communication equipment speech testing, this paper proposes a subjective and objective comprehensive evaluation method based on feature fusion. It focuses on extracting feature parameters such as range, variance, skewness, and kurtosis to characterize the transmission characteristics of speech signals. Through the automatic extraction of speech signal features and objective score calculation, and finally through a large number of subjective and objective joint tests, the mapping relationship between objective scores and subjective evaluation levels was analyzed, completing the classification of speech quality levels for objective evaluation.