Hidden Semi-Markov Models

where the class label $q_n^{(r)}$ is corresponding to the contextual vector ${\mathbf{f}}_n^{(r)}$.

8.4 Multichannel HSMM

Multichannel HSMM was proposed to model multiple interacting processes (Natarajan and Nevatia, 2007b). In contrast to the basic HSMM that a process has a single state at any instant, this extension generalizes the HSMM state to be a vector ${\mathbf{S}}_t=(S_t^{(1)},\dots ,S_t^{(C)})$ representing the states of multiple processes. Each of the states, $S_t^{(c)}$, is dependent on all ...