matlabHTK

Matlab/Octave Interface to HTK for Bioacoustic

---

MatlabHTK requires HTK (http://htk.eng.cam.ac.uk/) installed on system

MatlabHTK uses VOICEBOX speech processing toolbox MATLAB routines to read/write HTK binary files (http://www.ee.ic.ac.uk/hp/staff/dmb/voicebox/voicebox.html)

---

How to cite

Please use: Ranjard, L., Reed, B. S., Landers, T. J., Rayner, M. J., Friesen, M. R., Sagar, R. L. and Dunphy, B. J. (2016). MatlabHTK: A simple interface for bioacoustic analyses using hidden Markov models. Methods in Ecology and Evolution. doi:10.1111/2041-210X.12688

---

Function documentation

The calls to HTK functions are indicated.

train_HTK()

• Process manually annotated sound recording files (Raven, Praat or Audacity) and convert to format readable to HTK (.mlf)
• Create directories and configuration files for HTK parameters (config, hmmprototype, configvar.txt, dict.txt, gram.txt, wavlist.txt, trainlist.txt, uniksequencetxt.mat)
• Extract the annotated sound signals and compute the window sound parameter values (HCopy)
• For each sound category, train a HMM using HTK function (HInit, HCompV, HRest)
• Clean up temporary files

recognise_HTK()

• Check the hidden Markov model network (HParse, HSGen)
• For each file in the argument directory, perform recognition using the set of HMM (HVite)
• Convert recognition file to label file
• Save recognition result summary to a .csv file for postprocessing
• Clean up temporary files

compare_label()

• Compare the annotation similarity between two label files for the same sound file
• Conduct a randomization test to assess significance of the similarity score
• Write a confusion matrix for the first input label file to the second input label file

plot_label()

• Plot histogram of activity for a given input label
• Consecutive time windows are considered and the proportion of the total window duration annotated as the input label is plotted
• Output a csv file indicating the proportions of each time window annotated as the input label