The application is intended for a Public Broadcast Organisation, delivering regional Radio programs 24/7. Some hours of the day, non-stop music is broadcasted, other hours, a live program is presented by a DJ in the Radio Studio. For the live hours we want to add a 'Visual Radio' option, e.g. a video stream is shown on the channels website so that the listeners can have a look in the studio and watch the DJ presenting his/her program.

AI-activation

Before we activate a video stream, we want to make sure that a DJ is presenting a live program from the studio. It would make no sense to stream live images from an empty studio e.g. while non-stop or prerecorded programs are being transmitted.

 

By means of the Visual Radio application, the activation of the video stream can be automated and is not dependant on the individual DJ’s having to switch such a facility on and off.

Implementation

In order to realize this using STM32 Edge AI, we need to implement..

- taking a live camera image of the person behind the presentation desk

- encode a rtmp- or http-stream and stream it to a window on the website

- override option to enable/disable streaming manually.

Future enhancements

At a later stage we will investigate possibilities to enhance of the application with face recognition. This way we could, before making a stream public..

- determine the identity of the DJ presenting the radio program

- check if he/she has agreed in being presented in a video window on the website.

Face recognition
For the possible enhancement with face recognition, we would additionalyy need..

- a small local database of radio makers that have agreed to appear on the website

- face recognition of the person behind the presentation desk

- identifying if this camera image is free to be streamed to the website.

To be continued..

 The Visual Radio application is being built, using the STM32N6 Discovery Kit.