FM signal capturing system and voice recognition for the assistance of individuals with hearing impairments.
$ git clone https://github.com/SanchezCris/SDR-Automatic-Speech-Recognition.git
- Supports LimeSDR through GNU Radio.
- Supports 2 MHz as sample rate and demodulates a signal at 48 KHz
- FM demodulation of an RF signal.
- Wav to Text
- Threading
- Printing in terminal and web graphical interface
There is an example GNU Radio Companion (.grc)
flowgraph located at SDR-Automatic-Speech-Recognition/app/gnuradio/fm_receive_tcp.grc
. To use it, double click on this file and run it.
Example of GUI GNU Radio.
Example of printing on terminal.
To view the real-time transcription of the audio signal, a webserver is included. The webserver should be started after the GRC flowgraph. Before running the webserver, be sure to install its dependences.
- Open a terminal at
SDR-Automatic-Speech-Recognition/app/
. $ conda activate sdr
$ streamlit run webserver_tcp.py [ARGUMENTS]
- Open a web browser
- Browse to
localhost:8501
GNU Radio is a dependence for fm_receive_tcp.grc
. I recommend installing it with this Ubuntu image.
If using webserver_tcp.py
, you'll need to install the following Python packages.
$ conda activate sdr
$ pip install numpy
$ pip install scipy
$ pip install speechrecognition
$ pip install streamlit
Perform the tests in an open space, preferably without nearby appliances, with the antenna placed vertically.