A multi-stream ASR framework for BLSTM modeling of conversational speech

Keyphrases

• Multi-stream 100%
• Bidirectional Long Short-term Memory (BiLSTM) 100%
• Memory Models 100%
• Conversational Speech 100%
• Tandem System 40%
• Triphone Model 40%
• Recurrent Neural Network 20%
• Single-stream 20%
• Cosine 20%
• Long Short-term Memory Architecture 20%
• Stream Systems 20%
• Recognizer 20%
• Phoneme 20%
• Conversational Speech Recognition 20%
• MFCC Features 20%
• Memory Prediction 20%
• Multiple Streams Framework 20%

Engineering

• Long Short-Term Memory 100%
• Joints (Structural Components) 25%
• Single Stream 25%
• Recurrent 25%
• Recognizer 25%
• Data Stream 25%
• Lstm 25%

Computer Science

• Bidirectional Long Short-Term Memory Network 100%
• Data Stream 20%
• Neural Network 20%
• Speech Recognition 20%
• Memory Architecture 20%
• Recognizer 20%