MP-SENet: A Speech Enhancement Model with Parallel Denoising of Magnitude and Phase Spectra

Ye-Xin Lu, Yang Ai, Zhen-Hua Ling

Demo Page |

• In this paper, we propose a novel Speech Enhancement Network that explicitly enhances Magnitude and Phase spectra in parallel, dubbed MP-SENet.
• MP-SENet is the first speech enhancement model that realizes explicit phase estimation and optimization by using phase parallel estimation architecture and anti-wrapping losses.
• MP-SENet mitigates the compensation effect between the magnitude and phase by explicit phase estimation, elevating the quality of enhanced speech to a new level.

Towards High-Quality and Efficient Speech Bandwidth Extension with Parallel Amplitude and Phase Prediction

Ye-Xin Lu, Yang Ai, Hui-Peng Du, Zhen-Hua Ling

Demo Page |

• In this paper, we propose a generative adversarial network-based speech bandwidth extension (BWE) model with the parallel prediction of Amplitude and Phase spectra, dubbed AP-BWE.
• AP-BWE realizes high-quality speech BWE by explicit amplitude-phase estimation and multi-resolution amplitude-phase discrimination.
• AP-BWE realizes efficient speech BWE by using all-convolutional architecture and all-frame-level operations.

Multi-Stage Speech Bandwidth Extension with Flexible Sampling Rates Control

Ye-Xin Lu, Yang Ai, Zheng-Yan Sheng, Zhen-Hua Ling

Demo Page

• In this paper, we propose a multi-stage speech BWE model named MS-BWE, which can handle a set of source and target sampling rate pairs and achieve flexible extensions of frequency bandwidth.
• MS-BWE comprises a cascade of BWE blocks, with each block featuring a dual-stream architecture to realize amplitude and phase extension, progressively painting the speech frequency bands stage by stage.
• We adopt the teacher-forcing strategy to mitigate the discrepancy between training and inference.

Incremental Disentanglement for Environment-Aware Zero-Shot Text-to-Speech Synthesis

Ye-Xin Lu, Hui-Peng Du, Zheng-Yan Sheng, Yang Ai, Zhen-Hua Ling

Demo Page

• In this paper, we propose an Incremental Disentanglement-based Environment-Aware zero-shot text-to-speech (TTS) method, dubbed IDEA-TTS, that can synthesize speech for unseen speakers while preserving the acoustic characteristics of a given environment reference speech.
• IDEA-TTS is capable of environment-robust TTS, environment-aware TTS, and environment conversion with a single model.