AI-powered
podcast player
Listen to all your favourite podcasts with AI-powered features
Is It Possible to Detect Speech During the Preparation Phase?
The reason why we have the preparation phase is to shift any sort of uncertain, like, if you look at the neural responses before he tries to start speaking. And you don't want the detection model to confuse that with a real speech event until it's like really sure. So in some ways, it's an engineering trick to just try and get like very solid speech probabilities. In some other ways, it provides like a stepping stone for the model to If someone is trying, is about to speak, but actually doesn't end up quite getting there decides to like, you know, not speak quite yet, it would not go ahead and predict the full speech of them.