A Listener Rating Method for Network Subjective Listening of Speech Quality

Abstract

The invention relates to a comprehensive listener grade evaluation method for subjective audiometry and belongs to the direction of subjective evaluation of voice quality in the field of telecommunication technology. The method includes the following steps that 1, five listener data analysis indexes of subjective audiometry of voice quality are worked out; 2, quantification processing is conducted on the indexes through a hierarchical analysis quantification method so as to obtain the weight vector of each index; 3, a listener evaluation alternative set including excellence, goodness, medium, qualification and badness for online subjective audiometry experiments is established; 4, overall mean square values are calculated by combining the listener data analysis indexes with the weight vectors and divided into the listener evaluation alternative set. Compare with existing methods, the method has the advantages that performance of listeners participating in network audiometry experiments can be evaluated comprehensively, and therefore listeners with low evaluation grades can be rejected, and accuracy and credibility of network audiometry experiments are improved.

Description

technical field [0001] The invention relates to a method for assessing a listener's level, in particular to a method for comprehensively assessing a listener's performance level applied in the subjective listening of a voice quality network, and belongs to the direction of subjective assessment of voice quality in the field of telecommunication technology. Background technique [0002] The subjective evaluation of speech quality is an effective and important method to describe speech quality, and it plays an irreplaceable role in promoting the research and standard release of speech and audio codec technology. In order to ensure the scientificity, standardization and accuracy of the subjective speech evaluation, the International Telecommunication Union (ITU) has formulated a series of evaluation standards, but a large number of restrictions and strict requirements make the traditional speech subjective evaluation experiment a relatively inefficient and An expensive process,...

AI Technical Summary

Problems solved by technology

[0004] (1) The test environment is uncontrollable, and the listener may test in a noisy environment, which will obviously affect the accuracy of scoring

[0005] (2) The quality of listening equipment is uneven

In traditional audiometry experiments, high-quality playback devices and listening devices are provided, but for the listeners who use this system, since they use their own computers for audiometry experiments, their playback devices are computer sound cards , the listening device may be cheap headphones or amplified speakers of a personal computer, which will affect the accuracy of the experimental data as a whole

[0006] (3) Due to the lack of staff supervision and control, listeners may score some unreasonable scores because they cannot understand the audiometry test, or may try to cheat the system with random scores, which will make the system Received a lot of unreasonable experimental data, which adversely affected the experimental results

Images