This set of modules implement Audio Filtering, Speech Recognition, and Hotword Detection for the iCub.
The Audio Filtering Module is located in the src/audioFiltering folder. Audio Filtering uses the webrtc-audio-filtering library for beamforming/filtering. The script setup.sh can be used to download and install this library.
The module then publishes filtered audio on the port /filteredAudio:o
The folder src/PythonSpeechRecognizer contains multiple scripts for different forms of speech recognition.
This script performs basic Speech Recognition based on the model from the Tensorflow Examples. The recognized speech is then published on the port /speech:o.
This script performs large vocabulary speech recognition using Kaldi. The script needs a speech recognition server setup using docker.
The folder src/PocketsphinxSpeechRecognizer contains a module that performs speech recognition with PocketSphinx.