I made an online frequency analysis app

background

As a background of making a frequency analysis application this time, in the laboratory of the university I belong to, there are multiple groups studying pulse waves and electrocardiograms, but software for analyzing them ( The current situation is that there is only one PC with WaveLab paid software). Due to this situation, when one group is using the software, another group cannot be analyzed, so I decided to make it myself. Also, if frequency analysis can be performed online, it will be very convenient because it will be possible to perform analysis anywhere as long as you are connected to the Internet without having to install software.

About the language used

I used Python to create this web application. The reason is that I wanted to create it using Numpy and Scipy, which are famous libraries in Python. These two libraries have a lot of functions for frequency analysis, so we will create them using them. The reason I used Django in the framework was because I thought that Python's FW was Django, so I used it.

Creation of frequency analysis module

Create a module for analyzing pulse waves and electrocardiograms. The module name is analysis.py.

analysis.py

import io
import pathlib
import numpy as np
from scipy import signal

import matplotlib.pyplot as plt

from .models import Pulse_Rate

def setPlt(pk):
    #Get target data
    pulse_rate = Pulse_Rate.objects.get(pk=pk)
    path = pathlib.Path(pulse_rate.data.url)

    #Read data
    f = open(path.resolve())
    lines = f.read().split()

    # -----Variable preparation-----
    N = 4096
    lines_length = int(len(lines) / 2)
    if lines_length < N:
        N = 2048

    dt = float(lines[2]) - float(lines[0])
    pulse = []
    for num in range(N - 1):
        pulse.append(float(lines[num * 2 + 1]))
    # -------------------

    # -----Sampling frequency calculation-----
    t = np.arange(0, N * dt, dt)  # time
    freq = np.linspace(0, 1.0 / dt, N)  # frequency step
    # -----------------------------

    # -----Waveform generation-----
    y = 0
    for pl in pulse:
        y += np.sin(2 * np.pi * pl * t)
    # -------------------

    # -----Fourier transform-----
    yf = np.fft.fft(y)  #Fast Fourier transform
    # --------------------

    #Power spectrum calculation
    yf_abs = np.abs(yf)

    # -----Graph generation-----
    plt.figure()
    plt.subplot(211)
    plt.plot(t, y)
    plt.xlabel("time")
    plt.ylabel("amplitude")
    plt.grid()

    plt.subplot(212)
    plt.plot(freq, yf_abs)
    plt.xlim(0, 10)
    plt.xlabel("frequency")
    plt.ylabel("amplitude")
    plt.grid()
    plt.tight_layout()

A pulse wave or electrocardiogram graph and a power spectrum graph are generated like this. The data is read from the text file uploaded by the user.

Display a graph on a web page

In the module above, we are creating a graph with plot, so I would like to display this on a web page in svg format.

analysis.py

def pltToSvg():
    buf = io.BytesIO()
    plt.savefig(buf, format='svg', bbox_inches='tight')
    s = buf.getvalue()
    buf.close()
    return s

Now you can convert the plot to svg and display it on your web page.

views.py settings

Write views.py using the above module.

views.py

def get_svg(request, pk):
    setPlt(pk)       # create the plot
    svg = pltToSvg() # convert plot to SVG
    response = HttpResponse(svg, content_type='image/svg+xml')
    return response

This will return a response in HttpResponse (svg format).

urls.py settings

Set the URL.

urls.py

urlpatterns = [
    ...
    path('pulse_rate-detail/<int:pk>/plot/', views.get_svg, name="plot"),
]

templates settings

I would like to make the page that displays the graph this time a detail screen.

pulse_rate_detail.html

<img src="{% url 'pulse_rate:plot' object.pk %}" width=600, height=300>

You can now view the graph.

Created graph

The graph that can be actually created is as follows.

Generate filtered graph

The figure above is a graph of unfiltered data, so you can't see anything as it is. Therefore, I added a filtering function to generate a graph. (I will omit the code) When you apply the filter, a pulse-like waveform appears. (Since this pulse wave is output using the G component of the RGB value taken by the camera, it will not be a perfect waveform.) The lower and upper frequencies of the filter can be adjusted by yourself from the web page.

Summary

The application created this time will be used in the laboratory in the future. Also, in the laboratory I belong to, there are groups that are doing research other than electrocardiogram and pulse wave, so I would like to create something for those groups as well. In fact, my research theme is to measure a person's stress level from the temperature of the person's nose.

I want to make my laboratory the coolest laboratory in the university, so I want to create such an application even in the 4th grade!