band pass filter on 1s of data, returning a list of average amplitudes for each 100ms of data in a 1s sample.

band pass filter

import numpy as np
from scipy.signal import butter, lfilter, freqz, medfilt
from collections import namedtuple


def butter_bandpass(lowcut, highcut, fs, order=5):
    nyq = 0.5 * fs
    low = lowcut / nyq
    high = highcut / nyq
    b, a = butter(order, [low, high], btype='band')
    return b, a


def butter_bandpass_filter(data, lowcut, highcut, fs, order=5):
    b, a = butter_bandpass(lowcut, highcut, fs, order=order)
    y = lfilter(b, a, data)
    return y

# for 'low', 'med', and 'high' bands in daniel's thesis.
Band = namedtuple('band', ['low', 'high'])
bands = [Band(85, 222), Band(222, 583), Band(583, 1527), Band(1527, 4000)]

SAMPLERATE = 8000
#SPEAKING_THRESHOLD = 400

def filter_audio(data):
    # 85 to 583, playing around with constants in daniel's thesis
    signal = butter_bandpass_filter(data, 85, 583, SAMPLERATE)
    # median filter with a window length of 81 samples
    # each data array has 8000 samples, so we're doing 1/100 of the total length
    signal = medfilt(signal, 81)
    # # maybe zero values below a threshold?
    # zero_indices = signal < SPEAKING_THRESHOLD
    # signal[zero_indices] = 0
    return signal

def filter_amplitudes(data):
    """ we get 1s of data. split into 'samples' that are 1/10 second each, and calculate avg. amplitude.
    """
    filtered_signal = filter_audio(data)
    samplesPerAmplitude = 0.1 * 8000
    amplitudes = []
    sampleSum = 0
    for i in xrange(len(filtered_signal)):
        if (i % samplesPerAmplitude == 0 and i > 0):
            amplitudes.append(sampleSum / samplesPerAmplitude)
            sampleSum = 0
        else:
            sampleSum += abs(filtered_signal[i])
    if sampleSum != 0:
        amplitudes.append(sampleSum / samplesPerAmplitude)
    return amplitudes