C

Convtasnet Libri3Mix Sepnoisy

Developed by mpariente
ConvTasNet model trained on the Asteroid framework for noisy audio separation tasks, with training data from the Libri3Mix dataset.
Downloads 30
Release Time : 3/2/2022

Model Overview

This model is specifically designed for multi-speaker speech separation tasks in noisy environments, capable of isolating clear single-speaker speech from mixed audio.

Model Features

Efficient audio separation
Utilizes the ConvTasNet architecture to effectively handle multi-speaker speech separation tasks in noisy environments.
Optimized training configuration
Employs carefully designed filter banks and mask network configurations to enhance separation performance.
Comprehensive performance metrics
Provides multiple evaluation metrics including SI-SDR, SDR, SIR, SAR, and STOI for a thorough assessment of model performance.

Model Capabilities

Multi-speaker speech separation
Noisy audio processing
Audio enhancement

Use Cases

Speech processing
Meeting recording enhancement
Isolates clear single-speaker speech from multi-person meeting recordings to improve speech recognition accuracy.
SI-SDR improvement of 11.23, STOI improvement of 0.22
Voice communication denoising
Separates target speaker speech in noisy environments to enhance communication quality.
SIR improvement of 19.53
Featured Recommended AI Models
AIbase
Empowering the Future, Your AI Solution Knowledge Base
Š 2025AIbase