Limitations of surveillance lip reading—so stop believing all that is reported as “fact”.

Packard

Member
Joined
Nov 6, 2020
Messages
5,661
Location
Hudson Valley, NY
Lip reading depends on several factors for accuracy, mainly movement of the lips, context and history.

Some words have no lip movement, and some lip movement is associated with 2 or 3 different sounds.

For example “Mom”, “Bob” and “Pop” all look identical to the lip reader. Don’t believe me? Repeat those words in front of a mirror. And all could fit neatly in the same conversation.

Someone with history with the speaker might know that the speaker never used “Pop” for her father. Her go-to term was “Dad”. And that same someone would know that “Mom” was deceased. So the difficult word would fall to “Bob”. But the distant observer working for the news media is unlikely to have that intimate knowledge and the conversation would have a 1 in 3 likelihood of being correct. In other words, a guess.

Also, some words have no lip movement at all. Say “a dozen eggs” in front of the mirror if you don’t believe me.

Which brings me to the article that prompted this post:

MSM reported a conversation between Tiger Woods and Vanessa Trump.

And while I would have no problem reading Trump’s sentence by lip reading, Woods’ response is a problem.

“I don’t even have” is easily read. But “any hair” is like “a dozen eggs” and is unreadable—no lip movement. So, it is a guess. Was the guess made by the lip reader? Maybe. But I find it more likely that the article writer suggested it as it makes a better story than anything else he could think of.

"It's alright," as she confessed per the lip reader, "I know...I need to stop being immature."

Tiger Woods responded: "I don't even have any hair."

The MSM article link:


There are certainly better lip readers than I. But there are some limitations that cannot be bridged. So question every reported “lip read” conversation.

On the other hand some “read” conversations seem just as loud and clear as if the speaker was standing in front of you.

In the 1988 Olympics, Debi Thomas, an American figure skater, faced off against the (then) East German Katarina Witt.

I was watching the televised live performance on a TV screen in a TGI Friday restaurant. Thomas just finished her warm up and skated up to her coach and said, “I’m so tired”. There was no audio in the restaurant. This was pure lip reading.

She could have been standing next to me when she said that. All the words were easily read. I knew at that moment that she had talked herself out of a medal. I looked for someone who wanted to bet on the outcome of the competition (I know it was cheating) but I could find no takers.

Even a non-lip reader could read, “I’m so tired”. Check it out in front of a mirror.

This is Google’s summary:

Katarina Witt (East Germany) defeated USA's Debi Thomas to win gold in figure skating at the 1988 Calgary Winter Olympics, a showdown known as the "Battle of the Carmens" because both skated to Bizet's Carmen. While Thomas was a top contender, she struggled in her free skate with three jump errors, allowing Witt to secure her second consecutive Olympic gold.
 
Last edited:
I saw a story yesterday about a guy who was arrested because some facial recognition software said he was someone else. a 100% match, on another man, that was 4" shorter and 50# lighter. He was detained and handcuffed by the casino staff! while waiting on the police, who went along with it!
Plus the arrested guy had a CDL, which the other did not. Both were confirmed "Real ID" and they still insisted it was the same person?
 
I read about that too.

I’m guessing that with the advancements in facial recognition and A.I. that more of these mistakes will happen.

They are now using facial recognition and A.I. to dub in voices on TV shows. The show I saw had a “flash back” to a much younger version of the lead character. They cast a much younger, but similar looking actor. However, they dubbed in the current days’ actor’s voice.

I could always tell when they did this because invariably the lip movements were not a precise match to the dialogue. But in this case it was a perfect match. It was on my DVR and I played that scene a couple more times, and I never spotted a miscue.

My guess it was a combination of facial recognition and A.I. I found it to be impressive, but scary.
 
Back
Top