GithubHelp home page GithubHelp logo

mrkrd / cochlea Goto Github PK

View Code? Open in Web Editor NEW
108.0 19.0 40.0 5.32 MB

Inner ear models for Python

License: GNU General Public License v3.0

Makefile 0.01% Python 63.10% C 24.55% C++ 1.36% MATLAB 1.88% Batchfile 0.04% Cython 9.06%
python cochlea inner-ear model auditory computational-neuroscience neuroscience sound basilar-membrane inner-hair-cell

cochlea's Introduction

cochlea

cochlea is a collection of inner ear models. All 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
                                            (Auditory Nerve)

The package contains state-of-the-art biophysical models, which give realistic approximation of the auditory nerve activity.

The models are implemented using the original code from their authors whenever possible. Therefore, they return the same results as the original models. We made an effort to verify it with unit testing (see tests directory for details).

The implementation is 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 (Bio-Inspired Information Processing) at the TUM. It went through several versions and rewrites. Now, it is quite stable and we decided to release it for the community.

Features

  • State of the art inner ear models accessible from Python.
  • Contains full biophysical inner ear models: sound in, spikes out.
  • Fast; can generate thousands of spike trains.
  • Interoperability with neuron simulation software such as NEURON and Brian.

Implemented 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 and examples (probably the easiest is to start with run_zilany2014.py).

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()

You can browse through the API documentation at: https://pythonhosted.org/cochlea/

Installation

pip install cochlea

Check INSTALL.rst for details.

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:

index duration type cf spikes
0 0.15 hsr 8000 [0.00243, 0.00414, 0.00715, 0.01089, 0.01358, ...
1 0.15 hsr 8000 [0.00325, 0.01234, 0.0203, 0.02295, 0.0268, 0....
2 0.15 hsr 8000 [0.00277, 0.00594, 0.01104, 0.01387, 0.0234, 0...
3 0.15 hsr 8000 [0.00311, 0.00563, 0.00971, 0.0133, 0.0177, 0....
4 0.15 hsr 8000 [0.00283, 0.00469, 0.00929, 0.01099, 0.01779, ...
5 0.15 hsr 8000 [0.00352, 0.00781, 0.01138, 0.02166, 0.02575, ...
6 0.15 hsr 8000 [0.00395, 0.00651, 0.00984, 0.0157, 0.02209, 0...
7 0.15 hsr 8000 [0.00385, 0.009, 0.01537, 0.02114, 0.02377, 0....

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()}
    )
    

The library thorns has more information and functions to manipulate spike trains.

Contribute & Support

Similar Projects

Citing

Rudnicki M., Schoppe O., Isik M., Völk F. and Hemmert W. (2015). Modeling auditory coding: from sound to spikes. Cell and Tissue Research, Springer Nature, 361, pp. 159—175. doi:10.1007/s00441-015-2202-z https://link.springer.com/article/10.1007/s00441-015-2202-z

BibTeX entry:

@Article{Rudnicki2015,
  author    = {Marek Rudnicki and Oliver Schoppe and Michael Isik and Florian Völk and Werner Hemmert},
  title     = {Modeling auditory coding: from sound to spikes},
  journal   = {Cell and Tissue Research},
  year      = {2015},
  volume    = {361},
  number    = {1},
  pages     = {159--175},
  month     = {jun},
  doi       = {10.1007/s00441-015-2202-z},
  publisher = {Springer Nature},
}

Do not forget to cite the original authors of the models as listed in Implemented Models.

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. His work was supported by the General Federal Ministry of Education and Research within the Munich Bernstein Center for Computational Neuroscience (reference No. 01GQ0441, 01GQ0443 and 01GQ1004B).

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. The thesis entitled Computer models of acoustical and electrical stimulation of neurons in the auditory system can be found at https://mediatum.ub.tum.de/1445042

This work was supported by the General Federal Ministry of Education and Research within the Munich Bernstein Center for Computational Neuroscience (reference No. 01GQ0441 and 01GQ1004B) and the German Research Foundation Foundation's Priority Program PP 1608 Ultrafast and temporally precise information processing: Normal and dysfunctional hearing.

License

The project is licensed under the GNU General Public License v3 or later (GPLv3+).

cochlea's People

Contributors

jencke avatar knoguchi avatar mrkrd avatar

Stargazers

 avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar

Watchers

 avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar

cochlea's Issues

human group delay

Dear all,
First, thanks for sharing your code. I'm using it and it is really neat!
I have been using the model available via pip but I have also been looking at your code here.
What is not clear to me is why for the human_group_delay the beta value from Harte et al. 2009 is being divided by 2.
I can see in the code that the human_group delay is not being used (cat instead) but if this happens, then the beta value needs to be corrected.
Could I kindly ask you what group_delay is implemented in the available code in pip repository?
Kind regards,
Jaime

python 3 support

is cochlea supporting python 3 now? Its one of the remaining 2.7 packages in the audio domain (see here)

Problems importing _pycat?

First, thanks for this! I saw the announcement come across the auditory list, and have gotten the time to check it out. (Greetings from BU!)

I'm running into what is probably a configuration issue, so my apologies for what may be a stupid question.

I've set up anaconda 32-bit on a windows 7 x64 box, running python 2.7.x. I'm running cochlea under the debugger, using PyCharm as my IDE. I've cloned both cochlea and thorns, and they reside in C:/Projects/cochlea

Running examples\run_zilany2014.py results in the following:

"C:\Users\gvoysey\Anaconda\python.exe" C:/Projects/cochlea/cochlea/examples/run_zilany2014_rate.py
Traceback (most recent call last):
File "C:/Projects/cochlea/cochlea/examples/run_zilany2014_rate.py", line 42, in <module>
     import cochlea
File "C:\Projects\cochlea\cochlea\cochlea\__init__.py", line 30, in <module>
from cochlea.zilany2009 import run_zilany2009
File "C:\Projects\cochlea\cochlea\cochlea\zilany2009\__init__.py", line 28, in <module>
from . import _pycat
ImportError: cannot import name _pycat

I'm not sure why this may be. Any thoughts?

Cannot run demo, getting float>integer issue

I'm having an issue running the demos. I'm running the demo in a Jupyter Notebook, but when I get to generating a tone, I get this error, which seems to be coming from numpy's ramped_tone() function. Any chance you know how to get around this?

`
TypeError Traceback (most recent call last)
in
1 fs = 100e3
2 cf = 1000
----> 3 tone = wv.ramped_tone(
4 fs=fs,
5 freq=1000,

/Library/Frameworks/Python.framework/Versions/3.8/lib/python3.8/site-packages/thorns/waves.py in ramped_tone(fs, freq, duration, pad, pre, ramp, dbspl, phase)
169
170 if ramp != 0:
--> 171 ramp_signal = np.linspace(0, 1, np.ceil(ramp * fs))
172 s[0:len(ramp_signal)] = s[0:len(ramp_signal)] * ramp_signal
173 s[-len(ramp_signal):] = s[-len(ramp_signal):] * ramp_signal[::-1]

/Library/Frameworks/Python.framework/Versions/3.8/lib/python3.8/site-packages/numpy/core/overrides.py in linspace(*args, **kwargs)

/Library/Frameworks/Python.framework/Versions/3.8/lib/python3.8/site-packages/numpy/core/function_base.py in linspace(start, stop, num, endpoint, retstep, dtype, axis)
118
119 """
--> 120 num = operator.index(num)
121 if num < 0:
122 raise ValueError("Number of samples, %s, must be non-negative." % num)

TypeError: 'numpy.float64' object cannot be interpreted as an integer
`

examples/stats_tuning example does not work

I have installed cochlea on Ubuntu and trying to run stats_tuning.py. This gives a strange error as follows. I havn't changed any code.
ERROR STACK -
Traceback (most recent call last):
File "stats_tuning.py", line 41, in
main()
File "stats_tuning.py", line 24, in main
model_pars={'species': 'human'}
File "/usr/local/lib/python3.6/dist-packages/cochlea/stats/tuning.py", line 53, in calc_tuning
model_pars=model_pars
File "/usr/local/lib/python3.6/dist-packages/thorns/util/maps.py", line 387, in wrap
result = func(**kwargs)
File "/usr/local/lib/python3.6/dist-packages/cochlea/stats/threshold_rate.py", line 88, in calc_spont_threshold
silence = np.zeros(fs*tmax)
TypeError: 'float' object cannot be interpreted as an integer

BUG: Windows 10 installation error

Hi there,

I am am trying to install with standard pip command and getting an error that Microsoft Visual C++ 14.0 is required. This seems like an issue for two reasons: I have MS Build Tools installed (version 15), and from the installation instructions it sounds like I should be able to get binaries and there shouldn't be any need to build anyway.

Thanks for any guidance (and for maintaining this extremely useful tool!).

example does not work

I installed cochlea-master on Windows 8 with Anaconda(python2.7 64-bit). It works successfully in jupyter notebook to processing a generated sound as shown in usage website. Then I am trying to run the example by using a sound file but there is something wrong.

C:\Users\Alice\Desktop\cochlea-master\cochlea-master\scripts>python run_zilany2014 --hsr=100 --msr=75 --lsr=25 --cf=1000 --species=human --seed=0 --dbspl=60 tone.wav
Processing tone.wav
Traceback (most recent call last):
File "run_zilany2014", line 162, in
main(args)
File "run_zilany2014", line 155, in main
space
File "run_zilany2014", line 108, in convert_sound_to_mat_unpack
convert_sound_to_mat(**args)
File "run_zilany2014", line 72, in convert_sound_to_mat
sound_raw = wv.resample(sound_raw, int(f.samplerate), int(fs))
File "C:\ProgramData\Anaconda3\envs\py27\lib\site-packages\thorns\waves.py", l
ine 100, in resample
new_signal = dsp.resample(signal, len(signal)*new_fs/fs)
File "C:\ProgramData\Anaconda3\envs\py27\lib\site-packages\scipy\signal\signal
tools.py", line 2203, in resample
Y = zeros(newshape, 'D')
TypeError: 'float' object cannot be interpreted as an index

I tried to transform the datatype into int

waves.py
100: new_signal = dsp.resample(signal, len(signal)*new_fs/fs)
replaced by: new_signal = dsp.resample(signal, int(len(signal)*new_fs/fs))

but then there were some other asserting problems.
(like "assert sound.ndim == 1". I checked my data, in which the sound.ndim=2).

I am new to python and I 'm not sure why this may be. Could you help me with it?

ffGn function bug?

I am attempting to write a high speed version of the Zilany 2014 model. I noticed an inconsistency in the /cochlea/zilany2014/utils.py file.

The ffGn function calculates the fGN @ line 73 for H == 0.5. However, the return statement (which returns y) is only present in the else statement. In the case of (H == 0.5), the ffGn function would not return an array?

Although, digging through the code it seems that (in both the cochlea and original Zilany model) H will be hardcoded at 0.9 for this simulation.

Hope this helps.

--Nas

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. 📊📈🎉

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google ❤️ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.