I have been using this [code][1] posted in Github from [tryan][2], which is written in Python, to build some LTSAs. It is basically Triton's algorithm but in Python. Since it is an open source code you can maybe change the algorithm to output the product you want. I was wondering if it would not be useful for you, since the resulting output of the LTSA is an image, converting the LTSA images into numpy arrays in Python. It might be a very naive step but depending on what you need it could be useful.

This repository also has the the relevant Matlab/Octave code. 



  [1]: https://github.com/tryan/LTSA/blob/master/python/ltsa.py
  [2]: https://github.com/tryan
  [3]: https://bioacoustics.stackexchange.com/users/40/etgriffiths