Neural Networks for Speech and Sequence RecognitionSequence recognition is a crucial element in many applications in the fields of speech analysis and recognition, time-series prediction, control and signal monitoring. This book applies the techniques of neural networks and hidden Markov models to problems of pattern and speech recognition, using real-world examples throughout. Highlights include the incorporation of domain knowledge with learning from examples, the description of contemporary advances such as recurrent neural networks, hybrids with hidden Markov models, and a thorough but straightforward use of mathematics. Neural Networks for Speech and Sequence Recognition will prove valuable to researchers and graduate students alike. |
Contents
The BackPropagation Algorithm | 8 |
Integrating Domain Knowledge and Learning from Examples | 37 |
Sequence Analysis | 57 |
Copyright | |
6 other sections not shown
Common terms and phrases
activations algorithm analysis application approach approximation architecture associated back-propagation Bengio called capacity chapter character classification computed connected consider constraints context continuous convergence corresponding cost criterion delays density dependencies derivatives described discriminant distribution equation error estimation example experiments figure frame frequency function Gaussian given gradient gradient descent hidden layer hidden units hybrid important improve increase initial input knowledge learning likelihood machine Markov matrix methods mixture module neural networks nodes observation obtained optimization output output units parameters particular patterns performed phoneme points prior probabilities problem Processing recognize recurrent recurrent networks reduce represent representation respect segmentation sequence sigmoid signal simple single space speakers spectral speech recognition static structure task temporal training set transformation transition update variable vector vowel weights window