Vosk Speech Recognition: voskSpeechRecognition (Python API)

• Introduction
• Use
• Requirements
• Status
• Related projects

Introduction

voskSpeechRecognition module use Vosk Speech Recognition API in python. This module performs speech recognition using Kaldi speech recognition backend and converts to text. Also use YARP to send text detection by network. Also admits YARP source audio like input. This module also publish recognition results in YARP port. voskSpeechRecognition require models to perform the module. Some pre-trained models in english, spanish, chinese, russian, french, german, portuguese, greek, turkish, vietnamese are available in vosk models.

Documentation available on docs.

Use

voskSpeechRecognition requires audio like input.voskSpeechRecognition models should be located in voskSpeechRecognition/models/model-x, being x your selected language. Download vosk models and extract content in your model-x dir. Also configure language.ini with your x selected language.

The process to running the program:

1. Execute programs/voskSpeechRecognition.py, to start de program.

python3 speechRecognition.py

2. Connect recognition source.

yarp connect /voskSpeechRecognition/data:o /yourport/data:i

NOTE:

• Data results are published on /voskSpeechRecognition/data:o

Configure language

To configure speech recognition language model, language table is attached:

Language table:

Table 1. Language table

Language	x
Spanish	es
English	en
Chinese	cn
Russian	ru
French	fr
German	de
Portuguese	pt
Greek	gr
Turkish	tr
Vietnmaese	vn

Requirements

voskSpeechRecognition requires:

• Install YARP 2.3.XX+
• Install pip
• Install vosk:

pip3 install vosk

Possible errors:

vosk python version requirements:

• vosk require python 3.8+ to be used in Windows.
• vosk require python 3.5+ to be used in Linux.
• vosk require python 3.8+ to be used in Mac OS X.
• vosk require python 3.7+ to be used in Raspbian. (Raspberry also require to download and install .whl manually. vosk Raspberry version here

Tested on: windows 10, ubuntu 14.04, ubuntu 16.04, ubuntu 18.04, lubuntu 18.04 and raspbian.

Status

Related projects

• Alpha Cephei: vosk speech recognition