waveread.py

import wave
import math
import collections
import sys

# Program written October 2018 for Frequency Central by Jet Holt
# call like this:
# python thisprogram.py /path/to/sound.wav r s
# r = resolution, in bits. ie how much data per sample
# s = steps, in bits. ie how many samples. Actual samples is 2^s
# eg: 
# python thisprogram.py ./bassdrum.wav 10 12
# to output 4096 samples (4096 = 2^12) that have 10 bit resolution.
# Outputting every sample (ignoring s) is not supported by this program.

#Function to fix signedness
#Also converts stereo data to mono by summing and halfing
def getData(file):
	if file.getnchannels() is 2:
		my_bytes = file.readframes(1)
		left = int.from_bytes(my_bytes[:file.getsampwidth()], byteorder='little', signed=True)
		right = int.from_bytes(my_bytes[file.getsampwidth():], byteorder='little', signed=True)
		return (left/2) + (right/2)
	else:
		return int.from_bytes(file.readframes(1), byteorder='little', signed=True)

# Open Wav file
filename = sys.argv[1]
if filename.endswith('.wav'):
    filename = filename[:-4]
wav = wave.open(filename + ".wav",'rb')

# Resolution determined by cmd arg and wav sample width
resolution = min(int(sys.argv[2]), wav.getsampwidth()*8)

# Number of steps
steps = int(sys.argv[3])

number_bytes = math.ceil(resolution/8)
missing_bits = (number_bytes*8)-resolution

# Determine the minimum and maximum values of the wav file.
# (Ie what is the highest and lowest sample value?)
originalMax = 0
originalMin = 0
for i in range(wav.getnframes()):
	data = getData(wav)
	originalMin = min(data,originalMin)
	originalMax = max(data,originalMax)
wav.rewind()

if (originalMax > -originalMin):
	originalMin = -originalMax
else:
	originalMax = -originalMin

originalRange = originalMax - originalMin

# Todo: Assumes the 'rest' state is 0.
# We use this  
centerline = 0

# The next two parts, 'find the start' and 'find the end' are used to ignore silence
# (because we don't want to have to edit our wav files manually)
# We want to go through our wav file and start taking note of the data once the data 
# changes by 1% of the range above or below the centerline.
# We will record the frames where this change occurs, then backtrack by 3 frames
# (we backtrack to account for the 1% of data that we dropped)
percent = originalRange * 0.01

#Find the start
dist = 10
d = collections.deque(maxlen=dist)
startFrame = 0
#If value stays withing 1% of centerline for a duration, it hasn't started yet
for i in range(wav.getnframes()):
	data = getData(wav)
	d.append(data)
	average = sum(d)/dist
	if (data > average and data > average + percent) or (data < average and data < average - percent):
		startFrame = max(0, i - 3)
		break

#Find the end
dist=150
d = collections.deque(maxlen=dist)
endFrame = wav.getnframes()
for i in range(wav.getnframes()-1,0,-1):
	wav.setpos(i)
	data = getData(wav)
	d.append(data)
	average = sum(d)/dist
	if (data > average and data > average + percent) or (data < average and data < average - percent):
		endFrame = min(wav.getnframes(), i + 3)
		break

validFramesCount = wav.getnframes() - (startFrame + (wav.getnframes() - endFrame))
frame_interval = validFramesCount/(2**steps)

print(filename + " " + str(frame_interval))

limitMax = (2**resolution) - 1
limitMin = 0

# Get our data and manipulate it so that the highest values and lowest values
# from the sample match with the highest and lowest possible with our resolution
data = []
count = 0
for i in range(2**steps):
	wav.setpos(math.floor((i*frame_interval) + startFrame))
	value = getData(wav)
	rescale = math.floor((value - originalMin) / (originalMax - originalMin) * limitMax)
	data.append(rescale)
	count = count + 1

# Now write our data to a file
# I break after ever 16 samples to make it easier to read
table = open(filename + '.txt','w')

count2 = 0
while count2 < count:
	tablestr = ""
	for i in range(16):
		tablestr += str(data[count2])+", "
		count2 = count2 + 1
	table.write(tablestr+"\n")

table.close()