Speech recognition and facial recognition have come a long way.
Years ago, I was using Dragon NaturallySpeaking. At the time, it was considered the best product around but it still had lots of opportunities for errors. In fact, they had an extra program or piece of hardware where you could dictate to it when you were in meetings then return to your desk to load it in. I never thought that would be especially useful because I found that, if you didn’t edit your speech, immediately, some of the errors would be so far away from what you’d said that you’d be unlikely to remember what you initially trying to get at. I made a personal habit of speaking and editing all in one session. But if you get a cold or have a tooth repaired with Novocaine impairing your speech, all bets were off on how well it would understand you.
The years have gone by. Now, my cable TV service has a remote control with voice recognition. I was skeptical. However, regardless how soft a family member speaks into it, no matter how garbled what they say sounds or whether they’re mumbling, this remote control device does a terrific job with all our voices. It’s amazing how much the technology has changed over so many years.
I had desperately wanted to write an article for Scientific Computing World about speech recognition for applications such as Forensics labs but never obtained quite the right contacts with which to do that. Meanwhile, just by chance, I had an opportunity to speak with law enforcement officials, yesterday (before you ask, no, I wasn’t arrested – I was at a presentation), and what is really the coolest thing ever is their facial recognition systems.
Let me digress a bit: ten years ago or so I worked in a building that had facial recognition technology for its security. Of course, as a backup, they still needed a security desk for visitors AND for all those of us that the system never recognized.
In any case, the building’s system refused to recognize me about one-third of the time. I spent more time at the security desk than I ever had, before. It just wouldn’t recognize my face. The bottom line is this: the building hated me. 🙁
Okay, so that’s a bit far-fetched that the building “hated” me but, seriously, there was something going on, there. My theory was that not all faces can be read in quite the same manner and we actual did know that the position of the head affected the reading.
In fact, today, those factors probably have mainly been taken into account. We all watch CSI and laugh when the scientists put a sample in the GC and it just comes out with results. But with facial recognition, it seems to work about as well as it does on TV. It’s incredibly advanced.
Unfortunately, I do think I know which LIMS system this agency was running but did not have an opportunity to ask if or how this all fit with their system. That would be a great article for Scientific Computing World if I were still writing for them, but currently unlikely to happen.