Zero crossing rate (Zero Crossing Rate，ZCR) Refers to the number of times the voice signal passes through the zero point (from positive to negative or from negative to positive) in each frame. This feature has been widely used in the fields of speech recognition and music information retrieval, and is a key feature in the classification of percussive sounds.

• characteristic:

(1). Generally speaking, unvoiced (unvoiced sound) And the environmental noise ZCR are greater than voiced (voiced) sound）；
(2). Because the ZCR of unvoiced sound and environmental noise are similar in size, they cannot be distinguished by ZCR;
(3). In practice, the zero-crossing rate is often combined with short-term energy characteristics for endpoint detection, especially ZCR is used to detect the beginning and end of unvoiced sounds;
(4). Sometimes ZCR can also be used for rough fundamental frequency estimation, but this is very unreliable unless there is a subsequent refinement process.

In general, the greater the zero-crossing rate, the higher the frequency.

import math
import wave
import numpy as np
import pylab as pl
def ZeroCR(waveData,frameSize,overLap):
    wlen = len(waveData)
    step = frameSize - overLap
    frameNum = math.ceil(wlen/step)
    zcr = np.zeros((frameNum,1))
    for i in range(frameNum):
        curFrame = waveData[np.arange(i*step,min(i*step+frameSize,wlen))]
        #To avoid DC bias, usually we need to perform mean subtraction on each frame
        #ref: http://neural.cs.nthu.edu.tw/jang/books/audiosignalprocessing/basicFeatureZeroCrossingRate.asp
        curFrame = curFrame - np.mean(curFrame) # zero-justified
        zcr[i] = sum(curFrame[0:-1]*curFrame[1::]<=0)
    return zcr

# ============ test the algorithm =============
# read wave file and get parameters.
fw = wave.open('21.wav','rb')
params = fw.getparams()
print(params)
nchannels, sampwidth, framerate, nframes = params[:4]
str_data = fw.readframes(nframes)
wave_data = np.fromstring(str_data, dtype=np.short)
wave_data.shape = -1, 1
#wave_data = wave_data.T
fw.close()

# calculate Zero Cross Rate
frameSize = 256
overLap = 0
zcr = ZeroCR(wave_data,frameSize,overLap)

# plot the wave
time = np.arange(0, len(wave_data)) * (1.0 / framerate)
time2 = np.arange(0, len(zcr)) * (len(wave_data)/len(zcr) / framerate)
pl.subplot(211)
pl.plot(time, wave_data)
pl.ylabel("Amplitude")
pl.subplot(212)
pl.plot(time2, zcr)
pl.ylabel("ZCR")
pl.xlabel("time (seconds)")
pl.show()

See article:
http://ibillxia.github.io/blog/2013/05/15/audio-signal-processing-time-domain-ZeroCR-python-realization/