cochlea 1.1

Description

cochlea is a collection of inner ear models. The models are easily accessible as Python functions. They take sound signal as input and return spike trains of the auditory nerve fibers:

                         +-----------+     __|______|______|____
 .-.     .-.     .-.     |           |-->  _|________|______|___
/   \   /   \   /   \ -->|  Cochlea  |-->  ___|______|____|_____
     '-'     '-'         |           |-->  __|______|______|____
                         +-----------+
          Sound                               Spike Trains

The implemented models are state-of-the-art biophysical models and give realistic approximation of the auditory nerve activity.

They are implemented using the original code from their authors whenever possible. We wanted to have implementations that gives the same results as the original models. We also made an effort to verify the implementation with unit testing (tests directory).

The models are also fast. It is easy to generate responses of hundreds or even thousands of auditory nerve fibers (ANFs). It is possible, for example, to generate responses of the whole human auditory nerve (around 30,000 ANFs). We usually tested the models with sounds up to 1 second in duration.

I developed cochlea during my PhD in the group of Werner Hemmert at TUM. It went through several versions and rewrites. Now, it is quite stable and we decided to release it for the community.

cochlea could be used by:

• researchers doing computational neuroscience (combines well with NEURON and Brian),

• experimenters that require realistic input spike trains,

• researchers improving inner ear models,

• everyone interested in understanding how hearing works.

Models

• Holmberg, M. (2007). Speech Encoding in the Human Auditory Periphery: Modeling and Quantitative Assessment by Means of Automatic Speech Recognition. PhD thesis, Technical University Darmstadt.

• Zilany, M. S., Bruce, I. C., Nelson, P. C., & Carney, L. H. (2009). A phenomenological model of the synapse between the inner hair cell and auditory nerve: long-term adaptation with power-law dynamics. The Journal of the Acoustical Society of America, 126(5), 2390-2412.

• Zilany, M. S., Bruce, I. C., & Carney, L. H. (2014). Updated parameters and expanded simulation options for a model of the auditory periphery. The Journal of the Acoustical Society of America, 135(1), 283-286.

• MATLAB Auditory Periphery by Meddis et al. (external model, not implemented in the package, but easily accessible through matlab_wrapper).

Usage

Check our online DEMO!

Initialize the modules:

import cochlea
import thorns as th
import thorns.waves as wv

Generate sound:

fs = 100e3
sound = wv.ramped_tone(
    fs=fs,
    freq=1000,
    duration=0.1,
    dbspl=50
)

Run the model (responses of 200 cat HSR fibers):

anf_trains = cochlea.run_zilany2014(
    sound,
    fs,
    anf_num=(200,0,0),
    cf=1000,
    seed=0,
    species='cat'
)

Plot the results:

th.plot_raster(anf_trains)
th.show()

Spike Train Format

Spike train data format is based on a standard DataFrame format from the excellent pandas library. Spike trains and their meta data are stored in DataFrame, where each row corresponds to a single neuron:

The column ‘spikes’ is the most important and stores an array with spike times (time stamps) in seconds for every action potential. The column ‘duration’ is the duration of the sound. The column ‘cf’ is the characteristic frequency (CF) of the fiber. The column ‘type’ tells us what auditory nerve fiber generated the spike train. ‘hsr’ is for high-spontaneous rate fiber, ‘msr’ and ‘lsr’ for medium- and low-spontaneous rate fibers.

Advantages of the format:

• easy addition of new meta data,

• efficient grouping and filtering of trains using _DataFrame functionality,

• export to MATLAB struct array through mat files:

  scipy.io.savemat(
      "spikes.mat",
      {'spike_trains': spike_trains.to_records()}
  )

Please, check thorns for more information and functions to manipulate spike trains.

Requirements

• Python (2.7)

• Numpy

• Scipy

• Cython

• Pandas

• Matplotlib (optional, for examples)

• docopt (optional, for the command line scripts)

• thorns (optional, for examples and stats)

• matlab_wrapper (optional, for the MAP external model)

Note: On Windows you can install a Python distribution such as Anaconda or Python(x,y) to fulfill most of the dependencies.

Installation

Quick install:

pip install cochlea

Note: Windows binary packages (wheels) are coming soon. At the moment you need cython and a working compiler to install the package.

Issues and Bugs

https://github.com/mrkrd/cochlea/issues

Other implementations

• Carney Lab

• Matlab Auditory Periphery

• DSAM

• Brian Hears

Citing

If you are using this software in your research, please make a reference: Rudnicki, M. and Hemmert, W. (2014) Cochlea: inner ear models in Python, https://github.com/mrkrd/cochlea.

When you use any of the models, always cite the original publications describing the model.

Acknowledgments

We would like to thank Muhammad S.A. Zilany, Ian C. Bruce and Laurel H. Carney for developing inner ear models and allowing us to use their code in cochlea.

Thanks goes to Marcus Holmberg, who developed the traveling wave based model.

We are grateful to Ray Meddis for support with the Matlab Auditory Periphery model.

And last, but not least, I would like to thank Werner Hemmert for supervising my PhD.