Basic voiced speech gender detection

gender.py

#!/usr/local/bin/python3
import numpy as np
import scipy as sp
from pylab import *
import wave
import os
import struct

def wavData(f):
    freq = f.getframerate()
    numFrames = f.getnframes()
    numChannels = f.getnchannels()
    #print("Number of frames:", numFrames)
    #print("Sample width:", f.getsampwidth())
    #print("Sampling frequency:", freq)
    #print("Channels:", numChannels)
    data = f.readframes(numFrames)
    data = struct.unpack("<{0}{1}".format(numFrames * numChannels, 'h'), data)
    return (freq, np.array(data[::numChannels]))

def process(freq, signal):
    data = sp.fft(signal)[1:]
    data = np.abs(data)
    data = data / len(signal) * 2
    freqs = np.linspace(0, freq, len(data))
    max = np.argmax(data)
    #50 = x * freq / len
    #50 * len / freq = x
    min = int(np.floor(80 * len(signal) / freq))
    max = int(np.ceil(255 * len(signal) / freq))
    print(">>> min", min, "\t> max", max)
    
    maxArg = np.argmax(data[min:max]) + min

    result = maxArg * freq / len(signal)
    print(">>>>>>>>> dominating frequency:", result)
    return result


print("Start")
for fname in os.listdir("train"):
    if fname.endswith(".wav"):
        fname = "train/" + fname;
        print('Processing', fname)
        
        with wave.open(fname, "rb") as f:
            freq, frames = wavData(f)
        
            l = min(40000, len(frames))
            ret = process(freq, frames)
            if ret < 172.5:
                detected = 'M'
            else:
                detected = 'K'
            print("detected", detected, detected in fname)
            if not detected in fname:
                open(fname + "_" + str(ret) + "_" + detected + ".wrongresult", "w").close()
    
print("End")