How can I convert a numpy array (wav data) to int16 with Python?

Question

I have some data that I assume is wav data. If I use:

soundfile.write(filepath, my_data, samplerate)

, then it correctly writes a wav file. But I want to somehow convert that wav data to int16 as currently it's some sort of float:

[0.0018415  0.00730521 0.01155283 ... 0.10048427 0.09344029 0.08903081]

with

max 0.3002103
min -0.33075073

Is there some way for me to convert to int16 this without having to save and then read a file?

_as currently it's some sort of float_ Can you check what it actually is? You've given practically no information here, can you please be more specific? Where is the file coming from? What libraries are you using? — AMC, Dec 24 '19 at 02:27
It's a numpy array? What's the `.dtype`? Do you want to just change the type to int16, or do you want to scale the data as well? — AMC, Dec 24 '19 at 15:11
I want to scale the data to the same that would have happened if I read with soundfile.read(dtype=int16) — Shamoon, Dec 24 '19 at 15:18
Is the data currently in the range [0, 1[ ? You write that you _assume_ the data is in the wav format, can you be more specific? — AMC, Dec 24 '19 at 15:19
I assume it's `wav` because when I `soundfile.write` it as a wav, it works — Shamoon, Dec 24 '19 at 17:11
@Shamoon does `soundfile.read(dtype=int16)` do linear scaling or logarithmic? — Mad Physicist, Dec 24 '19 at 18:03
@Shamoon That doesn’t sound very scientific lmao. Is that all you know about the array? — AMC, Dec 24 '19 at 18:04
@Shamoon. Actually they do. The library will round but not scale — Mad Physicist, Dec 24 '19 at 18:24

score 4 · Accepted Answer · answered Dec 26 '19 at 16:35

4

Answer was simple enough:

my_data = (my_data * 32767).astype(np.int16)

answered Dec 26 '19 at 16:35

Shamoon

1 Answers1