Skip to main content
Figure 8 | BMC Bioinformatics

Figure 8

From: A novel, fast, HMM-with-Duration implementation – for application with a new, pattern recognition informed, nanopore detector

Figure 8

Average Viterbi Decoding Accuracy over 10 different trials (instances) of 10 k-length synthetic 3-level signal data, where all levels have identical Poisson duration but the separation (gaussian emission means) between the levels varies. The Viterbi decoding accuracy improves as the number of bins increases in the decoding HMM's approximation of the Poisson durations generated using a 1 k-bin length distribution representation in the generating HMM. From left to right in each plot, the Viterbi response improves as the separation of the 3 levels (emission means) increases. Top, decoding performance when all levels have identical attributes is random 3-way guessing, so the expected 3333 out of 10000 correct is observed in all cases. Bottom, decoding performance with distributions with means 19 (for geometric), 19.25 and 19.5 (with Poisson distributed dwell-times).

Back to article page