Full-reference Speech Quality Estimation With Attentional Siamese Neural Networks

Abstract

In this paper, we present a full-reference speech quality prediction model with a deep learning approach. The model determines a feature representation of the reference and the degraded signal through a siamese recurrent convolutional network that shares the weights for both signals as input. The resulting features are then used to align the signals with an attention mechanism and are finally combined to estimate the overall speech quality. The proposed network architecture represents a simple solution for the time-alignment problem that occurs for speech signals transmitted through Voice-Over-IP networks and shows how the clean reference signal can be incorporated into speech quality models that are based on end-to-end trained neural networks.

Stats

• citations13
• S2 citations—
• github stars0
• HF likes0
• heat score8.60
• arxiv keymittags2021full

Related papers

• Leveraging Native Language Speech For Accent Identification Using Deep Siamese Networks (2017)7.50
• Full Attention Bidirectional Deep Learning Structure For Single Channel Speech Enhancement (2021)0.00
• Attentivemos: A Lightweight Attention-only Model For Speech Quality Prediction (2024)3.58
• Quality-net: An End-to-end Non-intrusive Speech Quality Assessment Model Based On BLSTM (2018)15.62
• Non-intrusive Speech Quality Assessment Using Neural Networks (2019)13.74
• Stoi-net: A Deep Learning Based Non-intrusive Speech Intelligibility Assessment Model (2020)0.00
• Norefer: A Referenceless Quality Metric For Automatic Speech Recognition Via Semi-supervised Language Model Fine-tuning With Contrastive Learning (2023)0.00
• Multi-modal Hybrid Deep Neural Network For Speech Enhancement (2016)0.00