Every car voice assistant has the same blind spot: it can hear you but it can’t see you. Say “open that window” and the system has no idea which of four windows you mean, so you end up saying “open the rear passenger-side window,” a phrase no human has ever said to another human. Hyundai’s patent activity suggests the company wants to fix that by pairing what the car hears with where your finger is pointing.
The Building Blocks Hyundai Already Owns
Before getting to pointing and voice together, look at the pieces Hyundai and Kia already hold. On August 11, 2026, the USPTO granted the two companies Patent 12,705,930, filed in March 2023. It covers a camera-based interface that splits the image of an occupant into regions, measures how much of each region the hand fills, and uses that to recognize the gesture. The clever part is that it first determines which way the wrist is pointing and divides the image based on that orientation. In plain terms, the system figures out which way your hand is aimed before trying to read what it’s doing, which is exactly the problem a pointing interface has to solve first.
The same day, Hyundai and Kia were granted a patent tackling the other half of the problem: noise. Patent 12,703,225 describes a controller that checks whether cabin noise exceeds a reference level while the climate system is running, and whether a voice command has started or spoken guidance is scheduled. If so, it adjusts the climate control based on the target and cabin temperature. Anyone who has repeated a navigation address three times over a blower on max knows why that matters. A third patent granted that day, 12,703,368, covers an interface that watches the passenger, classifies them by type, and serves up a layout meant to cut task completion time.
Related Articles
- South Carolina Arrested 80 People At A Street Takeover — But The Real Punishment Is What Happens To Their Cars
- BYD’s Pakistan Factory Missed Its Deadline Again — And BYD Isn’t The One Who Should Be Sweating
This line of work goes back years. Hyundai Mobis, the group’s parts and electronics arm, holds gesture control patents claiming priority to Korean applications filed June 25, 2019. They describe a system that receives an occupant’s gesture and triggers vehicle functions according to a learned usage pattern.
How Pointing Plus Voice Actually Works
The best way to see the engineering is in a granted U.S. patent that spells it out in unusual detail. U.S. Patent 12,485,756, which claims priority to a Korean application filed March 9, 2023, describes a vehicle that combines hand-motion recognition with voice recognition. Its stated goal is to perform the function the occupant wants “quickly and accurately.” (The USPTO text retrieved for this story did not show the assignee.)
The geometry is the interesting part. The system detects the occupant’s hand in the cabin camera image and maps it onto a 16-point skeletal model covering the palm and finger joints. It then draws an imaginary line from the extended fingertip toward the menu screen. Where that line hits the screen’s plane is where you’re pointing. The selected item gets enlarged or highlighted so you know what the car thinks you mean, and only then does it turn the microphone on.
Then comes the part that runs against today’s big-language-model voice assistants. Once an item is highlighted, the patent says the occupant can trigger it with a simple word like “operate” or “on.” They can even just make a loud enough sound if they can’t speak. The system doesn’t need to understand the word, only to tell that the sound came from the occupant rather than road noise.
It does that with two microphones, one aimed at the occupant and one sampling the whole cabin. It triggers when the difference in sound pressure between them passes a threshold. A single-microphone version compares the sound at the moment of speaking against the noise just before it.
That’s a smart shortcut. Speech-to-text inside a moving car is hard. Detecting that someone said something while pointing at the music icon is much easier. The patent argues this approach avoids recognition errors from dialects and accents, and avoids the lag of making someone hold a gesture still to confirm it.
The same document ties all this to Level 4 autonomy, where a fixed driver’s seat and fixed screen may disappear. Menus could be projected onto a surface or into open space, and the list of controllable functions extends to driver-assistance settings such as lane keeping and smart cruise control.
None of This Is New, Which Is the Point
Pointing combined with voice is one of the oldest ideas in human-computer interaction. Hitachi filed for a U.S. patent in 1993 on a display system that accepted user commands through both voice and gesture input, with pointing coordinates as a core element. It’s now assigned to Google.
In the car specifically, a U.S. application filed in February 2013 described giving separate confidence scores to voice and gesture inputs. The command executes only when both clear their thresholds, and the system asks for confirmation when only one does. A patent now assigned to Waymo links predefined gestures, including pointing, to specific regions of the vehicle, tying what the hand does to where it’s aimed.
So the question isn’t whether pointing plus voice can be patented as a concept. It’s whether Hyundai can patent a specific way of doing it: figuring out who is talking, which hand belongs to that person, and what they’re aimed at inside a moving cabin full of people. That’s where the value, and likely the legal fight, would sit.
The Regulatory Gray Zone
Federal distraction guidance wasn’t written with this kind of interface in mind. NHTSA’s 2013 distraction guidelines for built-in electronics are voluntary and cover visual-manual tasks. They recommend designs that let drivers complete tasks with glances away from the road of 2 seconds or less and no more than 12 seconds total. They also recommend that tasks not require more than one hand.
Pointing at a projected menu is still visual-manual; your eyes go to the target and your hand does the selecting. But if the target is a physical object like a window, vent, or mirror, the glance may be shorter than hunting through a touchscreen submenu. NHTSA’s own framework said a later phase would cover auditory-vocal tasks. A system that blends pointing with a spoken trigger falls between those categories, and nobody has written rules for that seam yet.
What It Means for Owners and Buyers
For now, nothing. Patents protect ideas; they don’t commit anyone to building them. Automakers file far more of these than ever reach a showroom.
If this does reach production, owners should think about the hardware underneath. The systems in these filings depend on interior cameras aimed at occupants and on multiple microphones placed around the cabin. That raises two practical questions.
Related Articles
- Ford Versus DOT: What The Federal Filings Reveal About Its China Ties
- Watch: A Corvette’s Donut Show Created Its Own Smoke Screen, Sent a Honda Into a House
First, repair. Camera-based features usually need calibration after the parts holding them are replaced or disturbed. A headliner repair, mirror swap, or trim job could become a calibration job too.
Second, privacy. An interface that constantly watches your hands to decide whether you’re pointing is, by definition, a camera watching the cabin. Buyers who care about that should ask whether video is processed in the car or sent to a server. The 12,485,756 patent describes both on-board and server-based speech processing.
The best-case version is a car where pointing at a window and saying “open” just works, with no menu tree. The worst case is one more system that has to be taught, calibrated, and occasionally argued with. Hyundai’s patent trail shows it’s working hard on the first outcome. Whether it gets there will be decided in a production car, not a patent office.