HARMONI opensmile

The HARMONI opensmile is a module that uses openSMILE detection to extract speech features from audio. It subscribes to the microphone sensor to detect speech features.

You have to run this module into the harmoni_detectors container.

Usage

The following documentation refers to the opensmile request.

The API for opensmile has:

  • Request Name: ActionType: START (to start the opensmile detection)

  • Body: no body

  • Response: no response only State (int)

  • Request Name: ActionType: STOP (to stop the opensmile detection)

  • Body: no body

  • Response: no response only State (int)

Run the following commands :

roslaunch harmoni_opensmile opensmile_service.launch

Paramaters

Parameters

Definition

Value

rate_frame

frame rate

int; 1

output_dir

directory path where the output files are stored

string; “$(find harmoni_opensmile)/output/ “

subscriber_id

name of the subscriber id

string; “default”

sampling_rate

sampling rate of the microphone

int; e.g., 16000

Testing

The module can be tested using

rostest harmoni_opensmile opensmile.test

The mock audio used is harmoni_stt/test_data/hello.wav It can be changed using the param test_opensmile_input

References

Documentation