eBay filed a patent for lip-reading technology. According to the abstract, it envisions a computing device that can “read the lips” of a user to enter data (and to determine when to switch from one input type to a lip-reading session).
The patent filing explains that conventional techniques in entering data into a “digital assistant” can be problematic if the user doesn’t want a bystander to hear their voice, or if ambient noise is high making voice recognition not function properly:
“The advent of digital assistant technology has alleviated some of these challenges. However, a digital assistant also introduces new challenges. For instance, conventional techniques to enter data using a digital assistant can become compromised in some situations because use of the digital assistant typically involves audible speech. Thus, conventional digital assistant techniques permit any person, device capable of recording speech, or another digital assistant, within the vicinity of the user to also observe this speech. This is especially problematic when entering sensitive or confidential information. In addition, certain environmental conditions render some input techniques ineffectual. If ambient noise is high, for instance, voice recognition functionality may not function properly. Thus, these limitations may result in inefficient user interaction with the device as well as inefficient use of computational resources of the device due to repeated user interactions required to achieve a desired result.”
The device would use machine-learning technology to personalize the lip-reading capability for each user. An excerpt of the filing explains:
“Machine learning can then be used to train the model by the computing device to recognize a user’s specific speech, visual cues (e.g., lip, tongue, eye, chin, cheek, head, shoulder movements), facial expressions, hand gestures, and mannerisms, thus personalizing the computing device for the particular user and adding another level of security.”
You can read the full patent filing on the USPTO.gov website.