the number of states

[Jiahn K]

why do traditional HMMs for ASR tend to model each phone with three distinct states?
I think simply using loop would be enough to control the duration?

[Simon King]

The three states allow the model to capture sound changes within a phone – the “beginning”, “middle” and “end”. Any number of emitting states could be used, but 3 has been found to work well and is – as you correctly state – the most common choice.

[Author]

Posts