Signal processing

Though magnetocardiography (MCG) devices are designed to measure the magnetic signal related to the cardiac activitiy, many instrumental error signals and environmental noises are recorded at the same time. Signal processing techniques are used to emphasize or detect components we have an interest in measured datasets. MCGpy also provides common signal processing methods by scipy.signal and numpy.fft as class methods for convenience.

Dependency modules are:

scipy.signal.butter - Butterworth digital and analog filter design
scipy.signal.lfilter - Filter data along one-dimension with an IIR or FIR filter
scipy.signal.welch - Estimate power spectral density using Welch's method
numpy.fft.fft - Compute the one-dimensional discrete Fourier Transform

Time-domain filters

TimeSeries and TimeSeriesArray classes support instance methods that remove some unwanted components or frequency bands from a signal. Available methods are:

Methods	Description
`bandpass(lfreq, hfreq)`	Only frequencies in a frequency band are passed.
`lowpass(freq)`	Low frequencies are passed, high frequencies are attenuated.
`highpass(freq)`	High frequencies are passed, low frequencies are attenuated.
`notch(freq)`	Only frequencies in a frequency band are attenuated.

Each one can continuously apply to time-series datasets. Here is a simple example of applying time-domain filters for single-channel data.

>>> from mcgpy.timeseries import TimeSeries
>>> hdf_path = '~/test/raw/file.hdf5'
>>> data = TimeSeries(hdf_path, number=1)
>>> filtered_date = data.notch(60).notch(120).bandpass(1,400).lowpass(300).highpass(10)
  

Frequency-domain filters

Additionally, TimeSeries and TimeSeriesArray classes include instance methods to obtain frequency-domain information via the following methods:

Methods	Description
`fft()`	Converting a signal from the time-domain signal to a discrete Fourier transform (DFT) of a sequence in the frequency domain.
`psd(seglength, overlap)`	Converting a signal from the time-domain signal to a power spectral density (PSD) in the frequency domain.
`asd(seglength, overlap)`	Converting a signal from the time-domain signal to a acceleration spectral density (ASD) in the frequency domain.

Here is a simple example of applying frequency-domain filters for single-channel data.

>>> from mcgpy.timeseries import TimeSeries
>>> hdf_path = '~/test/raw/file.hdf5'
>>> data = TimeSeries(hdf_path, number=1)
>>> fft = data.fft()
>>> asd = data.asd(2,1)
>>> psd = data.psd(2,1)
  

Accosicated classes

Note that in addition to the TimeSeires associated classes listed below.

Classes	Description
TimeSeires	Dealing with a single time-series of a MCG dataset
TimeSeriesArray	Dealing with a multi-channel time-series array of a MCG dataset