Hidden Markov Model Trainer

Press the 'Record Mouse Movement' button or the 'r' key to begin a 10 second recording period. If you are happy with a recording, press the 'Add Data to Conditioning Stack' button to add it to the set of data your Hidden Markov Model will be conditioned with. Add as many recordings as you like to the conditioning stack. More data representing the desired behavior generally results in a more accurate model.

Pressing the 'Condition HMM' button will imprint the data onto your HMM, overwriting its previous model. After imprinting data on your model, you will receive notifications of how well your HMM recognizes every mouse movement series you record. Recognition is based on the amount of mouse movement per second detected. The recognition scale is normalized (i.e. adjusted to be more readable by humans) to roughly 0-100.

Note that the maximum mouse movement per second value is capped, which is why any movement above a certain threshold will return the same result. All mouse movement values are converted to a whole number between 0 and 19 for easier processing by your HMM. [hide instructions]

[show instructions]

Recently Watched Movies