ANN  Artificial Neural Network. The framework that is used in the hybrid HMM/ANN recognizer to compute local phoneme probabilities.
ASR  Automatic Speech Recognition.
CDHMM  Continuous Density Hidden Markov Model. Currently the most popular flavor of the HMM for ASR.
CSR  Continuous (automatic) Speech Recognition
CTT  Centrum för talteknologi (Centre for Speech Technology), based at KTH, Stockholm. 
DAG Directed acyclic graph
DFA  Deterministic Finite State Automaton. 
DP  Dynamic Programming. An computationally efficient scheme for solving certain optimization problems.
EM  Expectation Maximization. Statistical method for parameter estimation.
EUROSPEECH  European Conference on Speech Communication and Technology. 
FSA  Finite State Automaton. 
HMM  Hidden Markov Model. A statistical time-series model that forms the basis of most contemporary  state-of-the-art ASR systems.
ICASSP International Conference on Acoustics, Speech, and Signal Processing. 
ICSLP  International Conference on Spoken Language Processing.
IEEE  The Institute of Electrical and Electronic Engineers, Inc.
JASA  Journal of the Acoustic Society of America.
KTH  Kungliga Tekniska Högskolan (Royal Institute of Technology), Stockholm, Sweden. 
MAP  Maximum a posteriori. Optimization criterion for the estimation of statistical parameters.
MCE  Minimum Classification Error. General optimization criterion for statistical classifiers.
ML  Maximum Likelihood. Optimization criterion for the estimation of statistical parameters.
NFA  Non-deterministic Finite State Automaton.
SD  Speaker Dependent.
SI  Speaker Independent. 
STL-QPSR  Speech Transmission Laboratory - Quarterly Progress and Status Report, THM, KTH.
TMH Institutionen för tal, musik och hörsel (Department of Speech, Music and Hearing), KTH, Stockholm.
TIMIT  Speech database recorded and processed by Texas Instruments and Massachusetts Institute of Technology.