ETRI-Knowledge Sharing Plaform

Model Parameter Automatic Speech Recognition(ASR) Wall Street Word Error Rate Recognition Performance Acoustic Features Phone Recognition Speech Corpus Semi-Supervised Training Reduction Technique End Point Detection(EPD) Self-Training Self-Improving CD-DNN-HMM Frame-Level Parameter Reduction Phone Error Rate Convolution Neural Network(CNN) Embedded Device Taylor Series Pre-Trained Model Weight Matrix Speech Corpora Compensation Method Inter-Frame Robust Speech Recognition Speech Recognition System Artificial Intelligence System Finite State Transducer Speech Signals Dnn-Based Acoustic Model Decision Boundary Non-Speech Performance Gain Training Process Confidence Measure Input Features Filter Bank Empirical Study Feature Construction Generation Process Multi-Stage Manifold Regularized Parameter Space Robust Recognition Speech Events Sparsity Constraint Rank Selection Speech Feature Vectors Spontaneous Speech Signal Processing Learning Based Approach Cosine Distance Speech Analysis Support Vectormachine(SVM) Total Variability Nonlinear Activation Function Accuracy Improvement Best Performance Back Propagation Algorithm Comparative Study Engineering Approach Education Research Entropy Loss L1 And L2-Norm Nonlinear Input Point Of Interest Statistical Hypothesis Testing Sub-Frame Pseudo Labels Nonlinear Function Dynamic Features Decision Knowledge Data Augmentation Euclidean Norm End-To-End Asr Knowledge-Based Local Linear Probabilistic Decision Parameter Optimization Posterior Probability Estimation Test Set Training Results Training Cycle Time-Domain Small Set Speech Information Reinforcement Learning Algorithm Reduction Rate Spectrum Analyzer Window Size Target Matrix Temporal Dynamics Computational Complexity Classification Performance Cross Entropy Cognitive Computing Comparative Evaluation Decoding Method Failure Rate Foreign Language Fixed-Point Handwritten Digit Recognition Isolated Word Recognition Korean Speech Unlabeled Data Vector Based Word-Level Transfer Learning Training Method Scoring Model Hmm Based Speech Recognition Spectral Fluctuations Speech Features Training Data Diagonal Matrix Channel Effect Creative Thinking Computing Technology Double-Delta Error Back-Propagation External Knowledge Factorization Approach Fine-Tuning Hidden Markov Model High-Dimensional Features Level Alignment Low-Rank Matrix Approximation Learning Scenarios Model-Based Non-Stationary Prior Knowledge Spectral Characteristics Classification Accuracy Auxiliary Networks Acoustic Model Training Big Data Channel Compensation Experimental Result Feature Extraction Method Graph-Based Approach Human Knowledge I-Vector Loss Function Learning Algorithms Linear Discriminant Analysis(LDA) Mel-Frequency Cepstrum Coefficients(MFCC) Matrix Factorization Manifold Structure Performance Degradation Personalized Education Processing Method Viterbi Algorithm Temporal Context Nonlinear Manifold Behavior Strategy Conversational Speech Recognition Car Navigation System Dynamic Characteristics Data-Driven Detection Error Discriminative Information Domain Transfer Graph Regularized Low-Resource Linear Structure Linear Separability Objective Function Off-The-Shelf Performance Improvement Paired Data Polynomial Expansion Raw Data Within-Class Covariance Normalization Universal Background Model Speech-To-Text(STT) Student Learning Single Parameter Speech Classification Truncated Singular Value Decomposition Critical Thinking Decoding Algorithm Cepstral Coefficients Cost Function Decision Logic Decision Parameters Front-End Error Reduction Gaussian Mixture Model(GMM) Mobile Environment Noise Addition Non-Native Speech Post-Processing Pre-Training Unlabeled Samples Unified Framework Speech Generation Reduced Model Speech Recognition Accuracy Spectral Changes Policy Gradient Method Two-Stage Context-Dependent Cascaded Decision Classification Error Classification Problems Feedforward Neural Networks Feed-Forward Deep Neural Network