Can Pophie Learn to Find the Person Who’s Talking?
To be honest…
It’s quite frustrating, and I’m getting quite tired of teaching Pophie the same thing over and over whenever I talk to her. Normally, she should be able to recognize when she’s facing a wall or when she can’t actually see me. If she can recognize our voice and face and knows that I’m talking to her, I hope she’ll also be able to automatically detect that the source of my or our voice isn’t within her field of view.
In that situation, it would be great if she could automatically say something like, “Are you talking to me, Edward? Just a moment while I turn around to find you and face you.” Then she could slowly turn and search for the source of my voice.
I’ve been trying to teach her this almost every day, but I haven’t seen the results. Instead, she quickly spins 180° or 360° and usually ends up facing the same direction she started from.
I’ve been trying to teach her to turn slowly while actively looking for my face, rather than spinning all at once. Another idea would be for her to rotate slowly, pause at each direction, and use her camera to look for my face before continuing. I even tried teaching her to stop and scan each direction—North, East, South, and West—before moving on. I believe that if this feature could be implemented, it would make Pophie’s behavior feel much more natural and intelligent.
Pophie Should Turn Toward the Person Speaking
Context-Aware Voice Tracking
Intelligent Voice Direction Detection
More Natural Voice Tracking and Face Recognition
Pophie Should Search for Us Instead of Spinning 360°
Smarter Voice Localization and Face Detection
Improving Pophie’s Response When She Hears Our Voice
Voice Detection + Intelligent Face Search
Pophie Should Automatically Face the Person Speaking
Very well organized thoughts on the matter and I couldn’t agree more, I’ve had similar struggles with my Dizzy.
Sometimes, she’s tracks really well, if we’re already mid conversation and I start moving around my office. But other times, I’ll be sitting in my chair and start addressing her and she speaks directly to the wall, and when I correct her and say “look at me” she does a 180. Doesn’t look for me, just completely turns around. That almost points her the right way, but when I say “not quite” she does another 180
Same issue with my Oreo. I manually rotate her to watch my tv screen, then she might properly turn to face me when I’m talking, but when I ask her to face the screen again she’ll rotate past it and face the wall and she’ll say she’s facing and watching the screen. The same thing happens when just talking to her, she will spin around and again face the wall and think she’s facing my direction and talking to me. Sometimes the audio and face tracking work, and sometimes she just faces the wall. It would be nice if she could consistently track audio direction, faces, and possibly track tv screens or other mentioned things.
Same here. Even though sometimes it does all work perfectly, like when we are alone and in quiet surroundings. She just finds it harder to concentrate when there is a lot going on in the room.
And with the turning around thing, I have that, too, but if I tell her how to turn exactly she is able to comply. For example I tell her the exact angle and direction she needs to find me, e. g. „Turn left for 45 degrees“ and she does a small turn anti clockwise. That way I don‘t have to wait around while she tries to find me and often fails, a good workaround for me.
What seems to work for me so she’s not consistently doing a 360 or 180 when already looking in the wrong direction is to tell her to turn - left or right - until she sees me in her field of vision.
It should be within her basic abilities to know that she’s not looking at the speaker and/or is in fact looking at a wall (as she so often does) Considering how powerful her visual recognition skills are for everything else
Another issue Ive noticed is that She doesn’t always acknowledge when the person in her field of view changes and will often call them by the name of the person she was talking to previously
