Deep speech inpainting of time-frequency masks

Abstract

Transient loud intrusions, often occurring in noisy environments, can completely overpower speech signal and lead to an inevitable loss of information. While existing algorithms for noise suppression can yield impressive results, their efficacy remains limited for very low signal-to-noise ratios or when parts of the signal are missing. To address these limitations, here we propose an end-to-end framework for speech inpainting, the context-based retrieval of missing or severely distorted parts of time-frequency representation of speech. The framework is based on a convolutional U-Net trained via deep feature losses, obtained using speechVGG, a deep speech feature extractor pre-trained on an auxiliary word classification task. Our evaluation results demonstrate that the proposed framework can recover large portions of missing or distorted time-frequency representation of speech, up to 400 ms and 3.2 kHz in bandwidth. In particular, our approach provided a substantial increase in STOI & PESQ objective metrics of the initially corrupted speech samples. Notably, using deep feature losses to train the framework led to the best results, as compared to conventional approaches.

arXiv commentAccepted to InterSpeech2020

References in corpus (6)

• Adam: A Method for Stochastic Optimization
• Very Deep Convolutional Networks for Large-Scale Image Recognition
• Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift
• SpecAugment: A Simple Data Augmentation Method for Automatic Speech Recognition
• Empirical Evaluation of Rectified Activations in Convolutional Network
• Understanding Neural Networks Through Deep Visualization

Cited by in corpus (4)

• GACELA -- A generative adversarial context encoder for long audio inpainting
• Improving performance of real-time full-band blind packet-loss concealment with predictive network
• On Deep Speech Packet Loss Concealment: A Mini-Survey
• Word-level Embeddings for Cross-Task Transfer Learning in Speech Processing