 This is the density of communication satellites around the earth There are more cell phones on earth today than there are people the whole world is talking and even more people are listening today One person can talk and a billion can listen the human voice has tremendous outreach And now it's being used to commit That was a hoax call Fraud threats harassment propaganda what have you in all these crimes voice is forensic evidence and my team at Carnegie Mellon is Developing technologies for profiling humans from their voice for tracking these criminals and I'll explain what that means through this example So the United States Coast Guard gets many distress calls each year These are made a calls made over radio when someone's life is in danger when they get these calls They launch search and rescue missions They send out boats and helicopters and specially trained personnel and they go out and search Hundreds of square miles at sea. It's dangerous and it's very expensive Turns out there are a bunch of jokers out there who like to make hoax calls It's they put other people's lives in danger. It's a federal crime back in 2014. There was a serial hoax caller Who kept making hoax like that over several months? After analysis of his calls, these are very short. This is what I concluded this person is white brought up in America Probably the Northeast. He's about 175 centimeters tall. He weighs about 75 kilograms. He's about 40 years old He's not in any kind of trouble. He's not on a boat. He's in a warehouse of some kind His warehouse has many machines. His radio room has a glass wall. His equipment is homemade He's still learning how to use it. He's sitting on a metal chair on a concrete floor This is what I call profiling humans from their voice and how can we do it? Turns out we do it all the time we make judgments about other people from their voices all the time How many times have you heard a voice and thought this person is sad this person is old this person is happy He sounded like a man who had slept well and didn't owe too much money We do it so easily that we don't even realize it and I'll demonstrate that through an example on the next slide a Remarkable ride only here in America. I was born in tell me who said this I Think you got it you just the person's gender age Ethnicity social background state of mental health and physical health all in two seconds for Profiling we computationalized human judgments We make machines do what we do only much better and for this we home in on information at very very fine levels Micro levels in the speed signal we use AI to discover this information to measure this information to learn from it and then to predict from it and Speaking of information you ll IV an that's my friend Tom Sullivan speaking out his last name What do you see on top is the voice signal? What is below are the is a representation of the frequencies in his voice and the patterns as you see change as he speaks this kind of a representation is called a spectrogram and Here you see the spectrograms of three different people saying the word up Do you see how the one in the center is different from the ones on the sites the persons on the sides are healthy The person in the center has Parkinson's you can visually see the signs of Parkinson's Over here on the next slide. I have Hitler from 1934 Turns out back in 1934 you already had signs of signs of Parkinson's He was never diagnosed. It has only been suspected from his video evidence Here is another example. These are snippets of the voice of the Queen of England's Christmas broadcast 50 years apart The arrows point to the same word. Do you see how the lines have smeared with H? They've also moved down. This is some of what happens to our voice when we age and it's not just the voice It's also the sound in which it is embedded this the recording for instance in this example The reflections in a room have caused a smear in the spectrogram and by studying that Spectrogram I can tell that this guy was in a room with extensive glass panels when he was talking Your voice is unique to you. It has information that your fingerprints and DNA don't have It has information about you at the time of speaking about just about the state of your body and your mind and All other kinds of things including your surroundings in fact at Carnegie Mellon We are working on recreating the human persona in 3d from voice alone