sslf 0.1.5

A simple spectral line finder

Simple Spectral Line Finder

sslf is designed to be a dead-simple, effective and useful spectral line finder for 1D data. It utilises the continuous wavelet transform from scipy, which is a productive way to find even weak spectral lines.

Because there are many definitions for a spectral line, this package can not conceivably work in all cases. Here, I have designed sslf for Gaussian-profiled spectral lines. A big part of my work focuses on recovering weak signals in the noise, so sslf aims to identify lines, flag them, and subtract the non-zero bandpass around them.

Usage

With a 1D spectrum (spectral_data) and a range of scales to test (e.g. 1 through 20), statistically significant spectral peaks can be found with:

import numpy as np
from sslf.sslf import Spectrum

s = Spectrum(spectral_data)
s.find_cwt_peaks(scales=np.arange(1, 20), snr=6.5)
s.subtract_bandpass()

The flattened spectrum is then contained in s.modified, and peak locations at s.channel_peaks.

find_cwt_peaks can optionally take the wavelet to be used in the wavelet transformation (Ricker wavelet by default). subtract_bandpass is a little harder to explain; see more complicated examples of usage can be found in the notebooks directory.

Limitations / Future work

I don’t know how to optimally select a wavelet for a particular spectral line. The Ricker wavelet appears to work well for Gaussian-profiled lines, though. I don’t know how well this extends to something like a Lorentzian, but you can spectify the wavelet to find_cwt_peaks. I believe the wavelet transformation is the most expensive part of this software. Rather than using scipy, pyWavelets may be faster, but I need time to try it out.

Contributions and Suggestions

I created this software because I could not find anything in the community that functioned similarly. So, I would welcome contributions to make this software more generally flexible, to be suitable for a wide audience. It’s really simple, just look at the code!

Bugs? Inconsistencies?

Absolutely! If you find something odd, let me know and I’ll attempt to fix it ASAP.

Contact

christopherjordan87 -at- gmail.com

Installation

pip install sslf

or

• clone this repo (git clone https://github.com/cjordan/sslf.git)

• within the directory, run: python setup.py install --optimize=1

Dependencies

• python 2.7.x or 3.x

• numpy 1.8+

• scipy

• future