BlueCat Motors E-books > Acoustics Sound > Automatic Speech Recognition: A Deep Learning Approach by Dong Yu

Automatic Speech Recognition: A Deep Learning Approach by Dong Yu

By Dong Yu

This e-book presents a accomplished evaluate of the hot development within the box of computerized speech acceptance with a spotlight on deep studying types together with deep neural networks and lots of in their versions. this is often the 1st automated speech popularity ebook devoted to the deep studying method. as well as the rigorous mathematical therapy of the topic, the publication additionally provides insights and theoretical origin of a chain of hugely profitable deep studying models.

Show description

Read Online or Download Automatic Speech Recognition: A Deep Learning Approach PDF

Similar acoustics & sound books

Bandwidth Extension of Speech Signals

Bandwidth Extension of Speech signs offers dialogue on varied methods for effective and powerful bandwidth extension of speech signs whereas acknowledging the effect of noise corrupted real-world signs. The publication describes the speculation and techniques for caliber enhancement of fresh speech signs and distorted speech indications reminiscent of those who have passed through a band trouble, for example, in a cellphone community.

Communication Acoustics (Signals and Communication Technology)

Verbal exchange Acoustics bargains with the basics of these components of acoustics that are concerning sleek verbal exchange applied sciences. end result of the creation of electronic sign processing and recording in acoustics, those components have loved an important upswing over the last four many years. The ebook chapters characterize overview articles protecting the main suitable components of the sphere.

How to Gain Gain: A Reference Book on Triodes in Audio Pre-Amps

The 34 chapters of the second version of the way to realize achieve provide a close perception right into a assortment (54) of the most typical achieve generating, consistent present producing probabilities, and digital noise production of triodes for audio pre-amplifier reasons. those chapters additionally supply whole units of formulae to calculate achieve, frequency and section responses, and signal-to-noise ratios of convinced construction blocks built-up with this kind of vacuum valve (tube).

Extra resources for Automatic Speech Recognition: A Deep Learning Approach

Sample text

In Chap. 12 we describe DNN-based multitask and transfer learning with which the feature representations are shared and transferred across related tasks. We will use multilingual and crosslingual speech recognition as the main example, which uses a shared-hidden-layer DNN architecture, to demonstrate these techniques. In Chap. 13 we illustrate recurrent neural network, including long short-term memory neural networks, for speech recognition. In Chap. 14 we introduce computational network, a unified framework for describing arbitrary learning machines, such as deep neural networks (DNNs), computational neural networks (CNNs), recurrent neural networks (RNNs) including the version with long short-term memory (LSTM), logistic regression, and maximum entropy model, which can be illustrated as a series of computational steps.

36) i=1 t=1 After carrying out similar steps for Q 2 (θ |θ0 ) in Eq. 35 we obtain a similar simplification N N T −1 Q 2 (θ |θ0 ) = P(q t = i, qt+1 = j|o1T , θ0 ) log ai j . 37) i=1 j=1 t=1 We note that in maximizing Q(θ |θ0 ) = Q 1 (θ |θ0 ) + Q 2 (θ |θ0 ), the two terms can be maximized independently. That is, Q 1 (θ |θ0 ) contains only the parameters in Gaussians, while Q 2 (θ |θ0 ) involves just the parameters in the Markov chain. Also, in maximizing Q(θ |θ0 ), the weights in Eqs. 37, or γt (i) = P(q t = i|o1T , θ0 ) and ξt (i, j) = P(qt = i, qt+1 = j|o1T , θ0 ), respectively, are treated as known constants due to their conditioning on θ0 .

Efficient backprop. In: Neural Networks: Tricks of the Trade, pp. 9–50. Springer (1998) 15. : The expectation-maximization algorithm. IEEE Signal Process. Mag. 13(6), 47–60 (1996) 16. : Minimum phone error and I-smoothing for improved discriminative training. In: Proceedings of International Conference on Acoustics, Speech and Signal Processing (ICASSP), vol. 1, pp. I–105 (2002) 17. : A tutorial on hidden markov models and selected applications in speech recognition. Proc. IEEE 77(2), 257–286 (1989) 18.

Download PDF sample

Rated 4.40 of 5 – based on 22 votes