Paper List in 2020

文章目錄

• • End-to-end
  • Noise Speech Recognition
  • Code-Switching

End-to-end

1 .END-TO-END MULTI-SPEAKER SPEECH RECOGNITION WITH TRANSFORMER

key :transformer,overlapped speech recognition,neural beamforming, speech separation

2. STREAMING AUTOMATIC SPEECH RECOGNITION WITH THE TRANSFORMER MODEL

這篇文章主要介紹了利用time-restricted self-attention對transformer的流式解碼實作.

3.JOINT PHONEME-GRAPHEME MODEL FOR END-TO-END SPEECH RECOGNITION

Noise Speech Recognition

1. IMPROVED ROBUST ASR FOR SOCIAL ROBOTS IN PUBLIC SPACES

   key: public space speech,Signal to noise ratio

   這篇文章主要介紹公衆場合，低信噪比環境對語音識别的影響。文中提到不同聲學環境下的混響系數T60的值分别是多少，以及不同聲學環境下的信噪比估計。

2. One-Pass Single-Channel Noisy Speech Recognition Using a Combination of Noisy and Enhanced Features

   key: single channel process,feature combine, feature/sub-network-level com-

   bination, gating mechanism

   這篇文章主要介紹了一種将降噪特征與原始特征進行聯合使用的方法. 經過speech enhancement 方法處理後的語音,雖然人耳聽覺感官上音質變得清晰,但是在語音識别系統中,往往由于語音增強造成的信号失真,導緻識别率反而降低. 本文通過将降噪特征與原始信号特征進行融合的方法,可以有效提高模型識别率.

3. Deep Learning for Distant Speech Recognition
4. An Overview of Noise-Robust Automatic Speech Recognition
5. SINGLE- AND TWO-CHANNEL NOISE REDUCTION FOR ROBUST SPEECH RECOGNITION
6. INVESTIGATION OF MONAURAL FRONT-END PROCESSING FOR ROBUST ASR WITHOUT RETRAINING OR JOINT-TRAINING

7. IMPROVING NOISE ROBUSTNESS OF AUTOMATIC SPEECH RECOGNITION VIA

   PARALLEL DATA AND TEACHER-STUDENT LEARNING

8. Adversarial Feature-Mapping for Speech Enhancement
9. Speech Denoising with Deep Feature Losses
10. A FULLY CONVOLUTIONAL NEURAL NETWORK FOR SPEECH ENHANCEMENT
11. Bridging the gap between monaural speech enhancement and recognition with distortion-independent acoustic modeling

Code-Switching

1. Code-Switched Language Models Using Neural Based Synthetic Data from Parallel Sentences