Real-time Multichannel IIR Filter in Python

Project Overview

This is a IIR filter class which uses the scipy sosfilt implementation to achieve the fast computational speed and high stability. The filter is capable of sample by sample multichannel realtime filtering and able to switch between raw mode and filter mode interchangably. Because of the sosfilt is implemented in C, the performance is very efficient. This class can be used in multichannel signal filtering such as real-time EEG monitoring system.

Import

import IIR filter use the following import:

from iir import IIR

Create an instance

The constructor takes the num_channels and sampling_frequency as an input parameters:

iir_filter = IIR(
    num_channels=4,
    sampling_frequency=256 #Hz
)

Building pipeline

Filter can be cascading by using the add_filter method. By default, the method constructs a butterworth filter with the same parameter as scipy.signal.butter method:

# Add 6th order butterworth bandstop filter with cutoff at 45-55Hz
iir_filter.add_filter(
    order=6, 
    cutoff=(45,55),
    filter_type='bandstop'
)
# Add 2th order butterworth lowpass filter with cutoff at 30Hz
iir_filter.add_filter(
    order=2, 
    cutoff=30,
    filter_type='lowpass'
)

For other types of filter, using scipy.signal.iirfilter method to create sos coefficient array and add to the IIR filter using add_sos method.

# Build your own filter
sos = scipy.signal.iirfilter(N=17, Wn=[50, 200], 
    rs=60, btype='band', analog=False, ftype='cheby2', fs=2000,
    output='sos')

iir_filter.add_sos(sos)

Filtering

The samples array is arranged in dimension of (N samples x M channels).

raw_samples = [
    [CH1[0],CH2[0],...,CHM[0]],
    [CH1[1],CH2[1],...,CHM[1]],
    [CH1[2],CH2[2],...,CHM[2]],
    ...
    [CH1[N],CH2[N],...,CHM[N]],
]

We can filter raw signals directly by calling filter method.

filt_samples = iir_filter.filter(raw_samples)

The state of the filter is remembered, so the next samples can be filtered by calling filter method over and over.

filt_first_chunk  = iir_filter.filter(raw_first_chunk)
filt_second_chunk = iir_filter.filter(raw_second_chunk)
filt_third_chunk  = iir_filter.filter(raw_third_chunk)

Raw mode

When testing a system, sometimes we want to check the raw signal without filtering. By using set_raw_mode method, we can bypass filtering operation in filter method. This is useful when you want to observe the raw signal quality by simply adding one line of code.

#Get filtered output
filt_samples = iir_filter.filter(raw_samples)

#Set to raw mode
iir_filter.set_raw_mode(True)

#Get the output the same as input
raw_samples = iir_filter.filter(raw_samples)

Examples

You can try running an example.py file to see the signal and frequency content plot before and after filter.

python examples.py

Two channel raw signals are generated as a combination of amplitude and frequency of:

200 @0.01Hz
5 @8Hz
20 @50Hz
2 @80Hz

The filter specification in example.py script are:

iir_filter = IIR(
    num_channel= 2,
    sampling_frequency= 250
)
iir_filter.add_filter(order=6, cutoff=(45, 55), filter_type='bandstop')
iir_filter.add_filter(order=4, cutoff=40, filter_type='lowpass')
iir_filter.add_filter(order=4, cutoff=3, filter_type='highpass')

The results of filtering is shown in below figure:

Acknowledgement

This project is inspired by py-iir-filter repository of @berndporr

what-in-the-nim / real-time-iir-filter