roboflex.audio_sdl

Support for reading audio data using Simple Directmedia Layer.

System dependencies

Requires python development libraries, and numpy. To compile the c++, we suggest a virtual python environment with numpy installed.

Requires SDL to be installed. More than likely, it already is installed in your distro. But if not:

apt-get install libsdl2-dev

# or maybe on mac:

brew install sdl2

pip install

pip install roboflex.audio_sdl

Import

import roboflex.audio_sdl as ras

Nodes

There is only one: AudioSensor

# all parameters optional: below are the defaults
audio_sensor = ras.AudioSensor(
    capture_id = -1,
    channels = 1,
    sampling_rate = 48000,
    capture_samples = 512,
    format = ras.BitDepth.F32,
    name = "AudioSensor",
    data_key = "data",
    debug = False,
)

# must be started (like all sensors)!
audio_sensor.start()

# you can get the realized channels and sampling rate (might be different from what you requested):
audio_sensor.sample_rate
audio_sensor.channels

# you can print the available detected devices, through a static method
ras.AudioSensor.show_devices()

This sensor publishes a TensorMessage, with the audio data encoded into a tensor of shape (C, S), where C is num channels, and S is capture_samples, under the key <data_key>.

Other

Available BitDepths:

# wrapped from SDL_AudioFormat: https://wiki.libsdl.org/SDL2/SDL_AudioFormat

enum ras.BitDepth:
    S8,
    U8,
    S16LSB,
    S16MSB,
    S16SYS,
    S16,
    U16LSB,
    U16MSB,
    U16SYS,
    U16,
    S32LSB,
    S32MSB,
    S32SYS,
    S32,
    F32LSB,
    F32MSB,
    F32SYS,
    F32