Optimization Of Speaker Extraction Neural Network With Magnitude And Temporal Spectrum Approximation Loss
2019 · Chenglin Xu, Wei Rao, Eng Siong Chng, et al.
Abstract
The SpeakerBeam-FE (SBF) method is proposed for speaker extraction. It attempts to overcome the problem of unknown number of speakers in an audio recording during source separation. The mask approximation loss of SBF is sub-optimal, which doesn't calculate direct signal reconstruction error and consider the speech context. To address these problems, this paper proposes a magnitude and temporal spectrum approximation loss to estimate a phase sensitive mask for the target speaker with the speaker characteristics. Moreover, this paper explores a concatenation framework instead of the context adaptive deep neural network in the SBF method to encode a speaker embedding into the mask estimation network. Experimental results under open evaluation condition show that the proposed method achieves 70.4% and 17.7% relative improvement over the SBF baseline on signal-to-distortion ratio (SDR) and perceptual evaluation of speech quality (PESQ), respectively. A further analysis demonstrates 69.1% an
Authors
(none)
Tags
Stats
Related papers
- Improving Speaker Discrimination Of Target Speech Extraction With Time-domain Speakerbeam (2020)14.76
- Similarity-and-independence-aware Beamformer: Method For Target Source Extraction Using Magnitude Spectrogram As Reference (2020)2.26
- Target Speaker Selection For Neural Network Beamforming In Multi-speaker Scenarios (2025)0.00
- Speaker Adapted Beamforming For Multi-channel Automatic Speech Recognition (2018)5.84
- Multichannel Loss Function For Supervised Speech Source Separation By Mask-based Beamforming (2019)7.50
- Enhanced Neural Beamformer With Spatial Information For Target Speech Extraction (2023)2.26
- Single Microphone Speaker Extraction Using Unified Time-frequency Siamese-unet (2022)3.58
- Target Confusion In End-to-end Speaker Extraction: Analysis And Approaches (2022)9.59