Personalized Audio Quality Preference Prediction

TL;DR

This study develops a siamese network-based method to predict individual audio quality preferences by incorporating both audio features and subject information, achieving improved accuracy over baseline models.

Contribution

It introduces a novel personalized preference prediction model combining audio input and comprehensive subject data, outperforming previous approaches.

Findings

Overall accuracy increased from 77.56% to 78.04%.

Using full subject information improves prediction performance.

LDNet with PANNs' CNN6 encoder outperforms baseline models.

Abstract

This paper proposes to use both audio input and subject information to predict the personalized preference of two audio segments with the same content in different qualities. A siamese network is used to compare the inputs and predict the preference. Several different structures for each side of the siamese network are investigated, and an LDNet with PANNs' CNN6 as the encoder and a multi-layer perceptron block as the decoder outperforms a baseline model using only audio input the most, where the overall accuracy grows from 77.56% to 78.04%. Experimental results also show that using all the subject information, including age, gender, and the specifications of headphones or earphones, is more effective than using only a part of them.