Keras Speaker Recognition - How to check exactness?

[TonyB22]

Hello !

I am using Keras Speaker Recognition to train my speakers and then find by audio file . https://keras.io/examples/audio/speaker_recognition_using_cnn/

For example I have 3 speakers

1.Tom 2.John 3.Mark

When I use some audio file that belongs to one of this speakers then it works great . But what if speaks Bob ? (that was not trained) Then it says it's Tom.

How can I check exactness ? I want it to throw error like : This speaker not found .

Please help , how to be with none trained speakers.

Thanks !