Non-reference audio quality evaluation method and system based on audio signal property classification

Abstract

The invention provides a non-reference audio quality evaluation method and system based on audio signal property classification. The method comprises the following steps: 1, establishing a training model based on a reference audio quality evaluation model, and adopting machine learning to acquire the relation between the audio quality of audio signals of different types and a network parameter, namely a non-reference audio quality evaluation model; 2, in network transmission of the audio signal network, inputting the current packet loss rate, the current delay time and the audio signal type of current packet loss data into the non-reference audio quality evaluation model, and acquiring the current audio quality. According to the non-reference audio quality evaluation method and system based on audio signal property classification, quality evaluation is conducted on signals of different types through different quality evaluation relations, user experience can be reflected more truly.

Description

technical field [0001] The invention relates to the field of audio quality evaluation, in particular to a non-reference audio quality evaluation method and system based on audio signal characteristic classification. Background technique [0002] With the improvement of people's quality of life, people have higher and more diversified demands on audio. Mobile audio services such as mobile audio on demand, audio information and music on mobile phones are growing rapidly. In order to ensure the good quality of audio services for users Experience, audio quality evaluation is essential. Audio quality evaluation is divided into subjective quality evaluation and objective quality evaluation. Although subjective quality evaluation can better reflect user experience, it is time-consuming, laborious, and not suitable for real-time transmission. Objective quality evaluation is often used in practical applications. Objective quality evaluation is divided into reference quality evaluati...

AI Technical Summary

Problems solved by technology

[0003] The existing no-reference audio quality evaluations are all obtained by statistical network parameters, such as packet loss and delay, and the most representative one is the E-Model model, but none of them considers the difference in the user's perceived distortion of different types of audio signals

Images