Hidden Markov models for burst error characterization in indoor radio channels

Many digital communication channels exhibit statistical dependencies among errors. The design of error control schemes for such channels and their performance evaluation is simplified if appropriate generative models of the overall communication link are available. This paper presents a new class of generative models based on the interconnection of hidden Markov submodels parameterized by the Baum-Welch algorithm. The method has some resemblance to the well-studied problem of speech recognition of isolated words; however, in our approach, instead of dealing with words, one deals with error bursts, and the final goal is to generate bursts rather than to recognize words. The proposed model is particularly suitable for simulating error profiles with long bursts, as is often the case in indoor radio channels, where the error-free gaps inside a burst are heavily nonrenewal. The merits of the method are corroborated by applying the technique to two particular examples of indoor code-division multiple-access (CDMA) radio links.