文档介绍：A Tutorial on Hidden Markov Models and
Selected Applications in Speech Recognition
LAWRENCE R. RABINER, FELLOW, IEEE
Although initially introduced and studied in the late 1960s and In this case, with a good signal model, we can simulate the
early 1970s, statistical methods of Markovsource or hidden Markov source and learnas as possible via simulations.
modeling have e increasingly popular in the last several
years. There are two strong reasons why this has occurred. First the fin all^^ the reason why signal are
models are verv rich in mathematical structure and hence can form important is that they often work extremely well in practice,
the theoreticai basis for use in a wide ranne of a~~- and enable us to im~ortant~ractica~ svstems-ex.. u.
ond the models, when appliedproperly, wGrk ve6"we/l in practice prediction systems, recognition sys;ems, ideniification SyS-
for several important applications. In this paper we attempt to care-
etc'l in a very efficient manner'
fully and methodically review the theoretical aspects of this type
These are several possible choices for what type of signal
of statistical modeling and show how they have been applied. . to
selected problems in machine recognition-of speech. model is used for characterizing the properties of a given
signal. Broadly one can dichotomize the types of signal
models into the class of deterministic models, and the class
of statistical models. Deterministic models generally exploit
Real-world processes generally produce observable out- some known specific properties of the signal, ., that the
puts which can be characterized as signals. The signals can signal is a sine wave, or a sum of exponentials, etc. In these
be discrete in nature (., charactersfrom afinite alphabet, cases, specification of the signal model is generally straight-
quantized vectors from a codebook, etc.), or continuous in forward; all that is required is todeterm