I am trying to find the pitch using aubio, but I can't read in the audioData directly as float32

Question

Is there a way I can cast all the values as float32 per CHUNK of data? I tried this way it doesn't give me an error but then aubio isn't able to recognize any pitch. Also I can't cast directly as float32 or it becomes too sensitive for FFT I am using in a different part of the program.

CHUNK = 1024 
#need it to be float for pitch, but int for frequencies
#FORMAT = pyaudio.paFloat32
FORMAT = pyaudio.paInt32

CHANNELS = 1
RATE = 44100

p = pyaudio.PyAudio()

stream = p.open(format=FORMAT,
                channels=CHANNELS,
                rate=RATE,
                input=True,
                frames_per_buffer=CHUNK)


audioData = stream.read(CHUNK)
pDetection = aubio.pitch("default", CHUNK,
         CHUNK, RATE)
    #     
# Set unit.
pDetection.set_unit("midi")
pDetection.set_silence(-40)
pDetection.set_tolerance(.1)
 #trying to change data into float         
 for i in audioData:
     i = np.float32(i)

  pitchSamples = np.fromstring(audioData, dtype = aubio.float_type)
    # 
    pitch = pDetection(pitchSamples)[0]

Answer 1

aubio expects float audio samples, not integers. To convert short integers ([-32768, 32767]) into floats ([-1., 1.]), you can simply multiply them by 1./32768 ~= 3.0517578125e-05.

def short_to_float(x):
    return x * 3.0517578125e-05

For reference, a complete example using floats is provided in demo_pyaudio.py.