My First Week with Myo: Data & Dreams

TL;DR After a week, data shows that: the Myo is very accurate even when your hands are full, the battery lasts much longer than I thought it did previously, its still comfortable, and does not damage easily.

It has been a bit of a wild week. My previous article, Magic in a Box: My First Day with Myo broke the top 10 on Hacker News, caught the attention of Thalmic Labs (which was pretty cool and somehow unexpected), and was seen by over 11,000 people in the first 12 hours. I’ve gotten some great feedback. Thanks for all of that.

This article has two parts. The first is a data heavy look at the Myo from a user perspective. Lots of graphs, and a link to my accuracy dataset if you want to dig at it more. The second is a pie-in-the-sky use-case brainstorm.

What is Myo?

Myo is a arm band that makes it possible to control computers and devices around you with simple hand gestures from anywhere within Bluetooth 4 range. Thalmic Labs is thinking big and building a device that stands to enable a ton of people. Lots of other motivated and cleaver people are building software and devices around the Myo. Here is a picture of one:

A Myo on a motivated developer

Who am I?

I’m currently an independent software engineer, but have worked with the smart people over at Amazon.com, Limelight Networks, and Arizona State University. I’m writing a book about Docker, blogging, recording, and building some exciting projects on my own. At the time of this writing, my only affiliation with Thalmic Labs is that after seeing their demo in 2013, I signed up for their developer program. If you’d like to chat about this stuff or be notified when I publish my next article, follow me on Twitter. This is what I look like:

A very focused @AllInGeek

Part 1: Data First

I learned the value of letting readers make their own data driven conclusions while working at Amazon. Not sure how I missed that in my last article, but I won’t make that mistake twice. So, I’ve presented all the data first with graphs.

Daily Bootstrap Effort

Measured how many times I had to adjust or calibrate the Myo when I put it on in the morning.

I think a major factor in my desire to use a device like the Myo is the effort it takes just to get it to work. In my experience over the last week, the Myo goes on pretty easy, but there is a small learning curve. It can be frustrating at first. The first time you put it on it might detect almost everything that you do as the open hand gesture. Or maybe it will recognize everything well except for the Thumb to Small Finger gesture. Having messed around with it a bit, I learned that when this happens you simply need to slip it down your wrist a bit, rotate it slightly and try again. After a few tries, it should work great.

In this last week I noticed that factors such as moisture on my skin made it a bit more difficult to get right. But once it is set, it stays that way.

Accuracy

Now, please bear with me here. This section is long. I conducted a three day experiment to test the accuracy of the device under different conditions. I measured accuracy with three factors, two treatments each, plus time:

• Arm Position: Elevated vs Hanging
• Hand: Empty vs Holding Something
• Gesture Pattern: Repetitive Sequence vs Random

The following digests were generated from 3700 sample tests. If you’d like a copy of the data its available in CVS form on GitHub.

The overall accuracy of the experiment was as follows:

These overall numbers roll up every treatment. From this data we can see that interestingly, there was an accuracy improvement over time. I don’t think this device learns, so either I was getting better at making gestures that the device could pick up, or I was getting a better placement each day. The step improvement between Monday and Tuesday might suggest the latter.

Rolling up the impact of the three factors and their two treatments can be seen here:

It is interesting to see that Arm Position and Pattern were so close with each treatment. I had honestly expected to see some difference between a pattern that is highly repetitive and a random gesture stream. But these two were almost exactly equal.

Less surprisingly, data shows that having your hands full while trying to gesture can have a negative impact on the accuracy of the Myo. Perhaps the only surprise here is that the accuracy was over 90%. I mean, that could be a pretty big deal. In this test I filled my hand with my iPhone 5s. I figured that would be appropriate since most people that have a smartphone usually have it in their hands. This was particularly tricky with the open hand gesture, when my arm was at my side. How do you open your hand and not let what you’re holding drop? Well I worked it out.

The Full Hand by Gesture data is probably pretty plain to see. I’m not going to dive too much deeper into that. But at this point it is worth calling out a few things about the Empty Hand data. With a sample size of 1,850 experiments with an empty hand, its reasonable conclude that the Myo is wonderfully accurate if you have a decent placement on your arm. The gesture break down of this data does support the claims that it is less accurate when detecting the Thumb to Small Finger gesture. But at 98.33% I think this is still in the realm of acceptable performance.

Battery Life

Measured time to battery exhaustion under various work loads.

Let me say that this is why its important not to make judgements based on single data points. My previous six hour report was plain inaccurate. I’m not sure what happened. Maybe leaving it unplugged overnight drained more than I expected at the time. Maybe I unplugged it before it was fully charged. Who knows…

Thursday and Friday I put the Myo on, used my computer for maybe an hour and went about my day away from anything it could bind to. I knew that the Myo would go into standby if not worn, but I wondered what kind of drain to expect if you are wearing but not using it. Well, the answer is all day long. Too long. I wanted the thing to run out of juice. I just wanted to go to bed. Fresh off the tails of the six hour report, I expected it to die at between six and eight hours. Thursday it went for twelve. Friday I finally took it off to sleep at over thirteen. This was a bit of a shock.

Saturday was a travel day, and so I was not able to put it through its paces.

Sunday through Tuesday I was wearing it all day again for my accuracy tests. These were heavy work days. I was almost always on my computer, taking measurements every hour and writing in between. Sunday the Myo ran for twelve hours. Monday, I worked under a similar load. After five hours I took it off for the night but did not set it to charge. I picked it up again Tuesday. Unsure what to expect I put it on and went to work. It lasted an additional five hours before dying.

At this point I’m happy to take back the things I said about being hesitant to use it for a surgery. It really is possible to get more than a full day’s work out of this tool.

Comfort

Counted the number of times per day and the timestamps when I adjusted the Myo, or wanted to.

Well, the data was unanimous. I never adjusted it. I never wanted to. I put it on, and mostly forgot about it. Once you get used to having something on your forearm, it is comfortable, never itches, and applies a perfect amount of pressure to stay put without cutting off circulation.

Durability

Daily summary of damage… After that first Wednesday, there was no damage. I checked for scratches, scuffs, dents, bends, or other damage. When I wasn’t wearing it, I put it in my back pack with chargers, headphones, and other things. While I did not intentionally abuse it, I didn’t exactly baby it. When I was wearing it, I rested my arm on plenty of hard surfaces. I bumped it into things. But I never had a problem, and it looks as good today as it did on Wednesday.

The only thing that concerns me is the joint where the band comes together. At least on mine, the seam is raised just slightly. I don’t think this is from wear. Its likely just the way it was built. I haven’t noticed it worsening at all, and really its not bad. But if this device has a weak point, I’d suspect that this is it.

Again, What’s the Verdict?

I stand by my previous statements. This is a great tool. I’m impressed that there is really nothing about it that would prevent its success in a consumer market. But that is only half of the battle. Now we’ve got to build things that people will want to integrate with. The onus is on the engineering community to make this happen. Unless I can walk from room to room in my house, feeling like I can magically control every aspect of my environment, I won’t want to wear it. Consumers will need more than a compelling story to use this daily, they’ll need a bunch of them.

Part 2: Dreams So Real, they’ll be Published in a few Months

Last time I disappointed a few people by holding back on what I’m doing with the Myo, and what I think might be interesting to see. So, this week I wanted to over deliver a bit. To set context, I’d like to talk about a different device for a moment.

Why not Leap Motion?

Leap Motion is absolutely awesome. It is a related product that detects gestures and other neat things. I’ve got one and had lots of fun with it. A former team member, who is all about personal efficiency, used one to stream line his development environment at work. At this point it seems much higher fidelity than the Myo. Leap Motion is a great interface expansion tool.

However, it is highly context bound. It has to be plugged into your computer and it only works within a very specific range above the device itself. Using it will give you Gorilla Arm. The Myo is worn, and connects via Bluetooth 4. It has decent range, but more importantly, it moves with you from place to place. This is what makes it an augmentation technology.

An Observation

Unlike a few other wearable tech innovations, nobody seems overly disturbed by the Myo. I’ve been wearing this thing everywhere for a week and only one person seemed to notice. They asked if it was a bracelet. I replied, “nope” paid for my energy drink and walked away when it was clear that they lost interest. But, I mean, I haven’t been hiding. Being independent, I work from all sorts of public places. I think this is a positive indicator. Augmentation technology should strengthen our connectedness to the world around us, not distract us or everyone else from it. So, in that respect this device is a winner.

I Wanted to Use it When

In my last article I said that I was going to wear it all day and keep track of things that I wanted to be able to control. I’ve found limited success with that. I think it can be challenging to see problems that you are used to dealing with. So the following is a mix of observational suggestions and thought experiment. I tried to stay away from things that I’ve already seen in Thalmic Labs’ promotional videos. Flying drones, gaming, controlling music, and driving robots would be awesome. But its been talked about already.

The Grind

I work on a computer almost all day. That being the case, its easy to come up with all sorts of things that would be nice to shortcut with gestures.

• Web browsing. People are already working on this. Seems kinda obvious. After all, there are lots of people that would want to do other things with their hands while… reading Reddit, CNN, FoxNews, etc.
• Development tools integration. Geeks like to build things for geeks. Shortcutting background version control commands or triggering rebuilds would be simple enough to implement.

Classes, Conferences, and Community

• Ad hoc voting. I’m sure this has been done with phones. But in environments like classrooms, phones can be a bit distracting. When I was in college there would be an occasional multiple choice survey. We had these really crappy remote control type devices with really lame LCDs. They were slow, and unsophisticated but they worked. Supposing you had a hub that could bind to multiple devices, and identify the device owners, that might be a nice target for replacement.
• Auction house bidding. Take silent auctions to the next level. Forget the paddles or whatever. Bind a Myo to a bidder.
• Community arm wrestling. So silly, I might just build this for fun. Again, supposing you have multi-device binding, it would be fun to build some sort of game around gesture consensus. Imagine a room full of people. A big projected graphic of two arms wrestling. The more people making a fist and twisting right the stronger the right arm is. Likewise for the left arm and twisting left. You could say that each second that each person makes a gesture, they contribute a point to one arm or the other. The first arm to, who knows, like a hundred points wins. Simple to animate. Plain to build. Maybe fun.
• Assistance. Thinking about places where you’re not typically surrounded by tech, or its quiet, or your being served, or a library (yes, those are still around, you should swing by and try one out). You could use Myo to silently request assistance.

Internet of Things (IoT)

This seems like augmentation technology. So, with that in mind, I think you’d see quite a bit of power in integration with several small things.

• Home automation. Think light dimming, temperature control, ceiling fans, open/close windows, raise/lower blinds, or turning on the garbage disposal (that switch is never quite close enough).
• Channel Surfing. I cannot believe we still use plastic sticks called remote controls. Even the first 5 gestures Myo provides would be enough to replace the remote control. If we can’t get it done with the Myo, could TV manufactures please stop using IR remotes? I hate IR remotes.
• Messy Hands. If you like getting your hands dirty Myo is great. Think cooking classes, art classes, performing surgery. Actually, I’m not sure the fidelity is there to be useful for surgery robots yet. Generally, it is a pain to have to clean up if you know you’re just going to get dirty again. Think gesture controlled stove tops and ovens. Maybe even a gesture controlled faucet, so you can clean up without making something else dirty.

Working for Two

I think the most interesting applications would enable a single person to do the work of two. Consider this device a shortcut to solving distributed work.

• Turn two keys simultaneously. You’d need a dedicated gadget that you could mount to the small surface area around a keyhole. It would hold the key in place, and turn it when you made the appropriate gesture. The security implications of such a device might be interesting. At the very least its a thought that I’d expect to see in a spy movie. I didn’t hate this enough to list it below. I just think it seems cool.
• Loosen or retract a winch. I drive a Jeep. Wearing the Myo around got me thinking about ways that you could use it with high utility vehicles. With Myo integration on a winch, you could work with your winch a bit faster. Eliminating the need to walk back and forth from the thing you are winching to your Jeep.
• Adjust spotlights. Again, inspired by Jeeps, it might be nice to be able to aim your mounted spotlights without actually being in or near your Jeep. Think about making camp and needing a bit more light in a different direction. With this integration you could hunt around for whatever was breaking those branches in the dark while keeping hold of your spear.

An Idea that I Hated

• Leveling Artwork. Forget about hanging a painting or something big when you’re alone. Hang the thing and put some device on the top that integrates with Myo. Adjust the painting by simply making a fist and twisting your wrist. After I thought about this for more than a few seconds, I realized that it might be silly to build a device so specific. Its not like people adjust their artwork every day. And even if their paintings or pictures were always on the move, why not build a device that does that automatically. I mean, without any input. It just always levels a frame to gravity or surrounding surfaces. Eh. Moving on…

These are Just a Few

Don’t horde your ideas. Chances are that even if you are a developer, and have a Myo, and are working on somthing for it, you won’t have the time to build everything that comes to mind. Its better to put them out there. You never know someone else might like them and help.

This device could be revolutionary. Thalmic Labs has set the stage with the device, but the development community will be the ones to complete the vision.