Orthogonal gradient penalty for fast training of Wasserstein GaN based multi-task autoencoder toward robust speech recognition

Keyphrases

• GaN-based 100%
• End-to-end Speech Recognition 100%
• Wasserstein 100%
• Fast Training 100%
• Multi-task Autoencoder 100%
• Gradient Penalty 100%
• Orthogonal Gradients 100%
• Automatic Speech Recognition 66%
• Wasserstein Generative Adversarial Network 66%
• Network Applications 33%
• Recurrent Neural Network 33%
• Denoising 33%
• Feature Enhancement 33%
• Noisy Environment 33%
• Deep Network 33%
• Convergence Rate 33%
• Speech in Noise 33%
• Speech Features 33%
• Penalty Method 33%
• Convolutional Neural Network 33%
• Wasserstein Distance 33%
• CNN Model 33%
• Noise Features 33%
• RNN Model 33%
• L1 Loss 33%
• L2 Loss 33%

Engineering

• Robust Speech Recognition 100%
• Automatic Speech Recognition 100%
• Autoencoder 100%
• Convolutional Neural Network 100%
• Convergence Rate 50%
• Noisy Environment 50%
• Recurrent Neural Network 50%