Search code examples
pythonaudionumpystreamreal-time

Noise while altering audio data

I am playing audio with python and i don't understand why i hear noise on the ouptut when executing code like this: 

import pyaudio  
import wave 
import numpy as np

f = wave.open('blabla.wav',"r")  
p = pyaudio.PyAudio()  
# open stream  
stream = p.open(format = p.get_format_from_width(f.getsampwidth()),  
                channels = f.getnchannels(),  
                rate = f.getframerate(),  
                output = True)  
float_array = np.fromstring(f.readframes(10000000), dtype=np.uint16).astype('float32')
output = 0.9 *  float_array
stream.write(output.astype('uint16').tostring())

When i multiply by 0.9 i expect weakening a signal a little bit. But where from this ouput noise came from ? I don't even add anything to initial data!

Basically i want to add two signals : 

   output signal =  0.5 * the origin one  + 0.5 * shifted origin one

But i get mess out of this process, because even a multiplying an original array make the signal near entirely sound like a mess.

Can you point me out what am i doing wrong and how to make the formula 

   output signal =  0.5 * the origin one  + 0.5 * shifted origin one

work right ?

Solution

  • I think 16 bit PCM is usually signed. Try using int16 instead of uint16