Uni ed Acoustic Modeling using Deep Conditional Random Fields
Journal Title: Transactions on Machine Learning and Artificial Intelligence - Year 2015, Vol 3, Issue 2
Abstract
Acoustic models based on Deep Neural Networks (DNNs) lead to sig- ni cant improvement in the recognition accuracy. In these methods, Hid- den Markov Models (HMMs) state scores are computed using exible dis- criminant DNNs. On the other hand, Conditional Random Fields (CRFs) are undirected graphical models that maintain the Markov properties of HMMs formulated using the maximum entropy (MaxEnt) principle. CRFs have limited ability to model spectral phenomena since they have single quadratic activation function per state. It is possible and natural to use DNNs to compute the state scores in CRFs. These acoustic models are known as Deep Conditional Random Fields (DCRFs). In this work, a variant of DCRFs is presented and connections with hybrid DNN/HMM systems are established. Under certain assumptions, both DCRFs and hybrid DNN/HMM systems can lead to exact same results for a phone recognition task. In addition, linear activation functions are used in the DCRFs output layer. Consequently, DCRFs and traditional DNN/HMM systems have the same decoding speed.
Authors and Affiliations
Yasser Hifny
Keywords
Related Articles
- EP ID EP278889
- DOI 10.41738/tmlai.32.1124
- Views 75
- Downloads 0
Body Temperature Patterns in Captive Barbary Striped Grass Mousse (Lemniscomysbarbarus): Influence of Photoperiod and Sex Steroids
The Barbary striped grass mouse (Lemniscomys Barbarus) has been described as a diurnal and photoperiodic rodent, but no reliable data exist on the daily and seasonal rhythmicity of body temperature (Tb). In this study, T...
Improved HMM for Cursive Arabic Handwriting Recognition System Using MLP Classifier
Recognizing unconstrained cursive Arabic handwritten text is a very challenging task the use of hybrid classification to take advantage of the strong modeling of Hidden Markov Models (HMM) and the large capacity of discr...
Data Analysis Application to Investigate Relationships between Metacognition and Learning Styles
The purpose of this study is to examine if learning styles predict effectiveness in learning. Participants of the study consisted of 80 students selected from different classes of the university of science and Technology...
Mitigating Economic Denial of Sustainability Attacks to Secure Cloud Computing Environments
In cloud computing environment where the infrastructure is shared by millions of users, attackers have the opportunity to ensure more damage to the compromised resources. The main aim of such attacks is to saturate and o...
Fuzzy Robust H Tracking Control For Wind Generator System: LMI approach
This study concerns the tracking control problem of the wind turbine generator system with uncertainties parameters and external disturbances. Based on TS fuzzy model, a fuzzy observerbased and a fuzzy robust state feedb...