Pure Mathematics
Baum, L., and T. Petrie. 1966. Statistical inference for probabilistic function of finite state Markov chains. Annals of Mathematical Statistics 37. 1554–63.
| The very first paper out of five published by Baum that describes the theory behind hidden Markov model. This paper requires background in probabilistic theory following mathematical publication tradition it begins with the traditional mathematical formulation: Let ... be ... without any abstract or introduction. It states and proves theorems related to how probabilistic transition distribution and observations probabilistic distribution can be recovered from observation sequences assuming that they are unknown, which is mathematically very beautiful.
|
Baum, L., T. Petrie, G. Soules, and N. Weiss. 1970. A maximization technique occurring in the statistical analysis of probabilistic functions of Markov chains. Annals of Mathematical Statistics 41. 164–71.
| Fourth paper from Baum's well-known series, it is centered on a theorem that stipulates that a conditional distribution based on past events will be attracted to a a limit distribution assuming an initial recurrent random walk. |
Speech Recognition
Rabiner, L. 1989. A tutorial on hidden Markov models and selected applications in speech recognition. Proceedings of the IEEE.
| This paper appears to be probably the reference on hidden Markov model theory with about four thousand citations. Readers don't need to have any background in statistical or probabilistic theory to fully understand the concepts explained in this paper since they are very well explained by the author. An extended bibliography on speech recognition is provided by the author as well.
|
Music recognition and classification systems
Pikrakis, A., and S. Theodoridis. 2005. A novel HMM approach to melody spotting in raw audio recordings. Proceedings of the International Symposium on Music Information Retrieval.
| This paper presents an interesting approach for melody spotting based on Variable Duration Hidden Markov Models. Observation staying at a given state for a long period of time is problematic for HMM.
|
Chai, W., and B. Vercoe. 2001. Folk music classification using hidden Markov models. Proceedings of the International Conference on Artificial Intelligence and Symbolic Computation.
| This paper presents a music classification system using HMM. Moreover the system includes rythm variation within the Markov Process. Rhythm dimension of melody is too often not taken in consideration in melody classification.
|
Adriane, D. 2001. Melody spotting using hidden Markov models. Proceedings of the International Symposium on Music Information Retrieval.
| This paper proposess a method for melody-based retrieval that adapts speech recognition techniques to melody spotting tasks. The implementation of the HMM is based on the left-right algorithm.
|