pcm2float

 I would like to share different implementations of Pulse-code modulation (PCM) to float conversion. Usually we receive PCM in int16 type, but some models expect normalized float64. The logic for conversion is not comlex and even so I have managed to find so interesting options.

The basic one logic looks like this:

 def pcm2float(sig, dtype=np.float64):s
    sig = np.asarray(sig)  # make sure it's a NumPy array
    assert sig.dtype.kind == 'i', "'sig' must be an array of signed integers!"
    dtype = np.dtype(dtype)  # allow string input (e.g. 'f')

    # Note that 'min' has a greater (by 1) absolute value than 'max'!
    # Therefore, we use 'min' here to avoid clipping.
    return sig.astype(dtype) / dtype.type(-np.iinfo(sig.dtype).min)

The code is taken from StackOverflow answer on a question SciPy wavfile: music in, garbage out?

What it does?

I think the most important is in the line. Let's split it on the logical parts:

- Convert NumPy array to desired dtype:

sig.astype(dtype)  # Converts numpy array to desired dtype

- Get machine limits for integer type

np.iinfo(sig.dtype)  # Returns object which looks like iinfo(min=-32768, max=32767, dtype=int16)

- Get largest value from limits object. Usually largest by module is minimal value:

-np.iinfo(sig.dtype).min  # -(-32768)

And that's pretty it. Because we used type to type conversion in line #4 we can convert minimal limit to required conversion type.

There is also more interesting variation of the normalization. I have found it in the python-audio repository Code example below:

    i = np.iinfo(dtype)
    abs_max = 2 ** (i.bits - 1)	# i.bits returns actual dtype number, e.g. np.int16 --> 16
    offset = i.min + abs_max
    return (sig * abs_max + offset).clip(i.min, i.max).astype(dtype)

Similar variation I've found in audio-format.py repository:

    i = np.iinfo(sig.dtype)
    abs_max = 2 ** (i.bits - 1)  # i.bits returns actual dtype number, e.g. np.int16 --> 16
    offset = i.min + abs_max
    return (sig.astype(dtype) - offset) / abs_max

Good examples can be found also in the next Jupyter notebook Reading and Writing Audio Files with wave.