Speaker Recognition not working correctly when I use different device for training and testing, Can you please help