Emotiv Insight Setup Guide

About

The Emotiv Insight is a Bluetooth device that you wear on your head to use facial expressions and cognitive commands to control your iOS device. Using this device with Brain Fu, you will be able to play a round of trivia using winks, smiles or even your own thoughts. That's right, a thought powered input device! It's kind of tricky to setup so I wrote this helpful guide to help get the most out of your Emotiv Insight experience with Brain Fu.

Connecting

Go to the Settings app on your phone, click on Bluetooth and make sure Bluetooth is enabled. When you launch Brain Fu it will automatically try to connect the Emotiv Insight. If it does not connect right away, you can manually initiate a connection by going to the Settings tab and clicking on the Emotiv option near the bottom of the list.

A banner will drop down from the top when the headset is connected. If this banner never appears, make sure the headset is turned on and double check that you have bluetooth enabled on your phone. Also make sure to close any apps in the background that might already be connected to the Emotiv headset.

Login to the Cloud

It is strongly recommended to store training data in your Emotiv cloud profile. This lets you train the headset once and re-use the same training data in every app. Once you are on the Emotiv setup screen and the headset is connected, click on the cloud button in the bottom right. Fill out your Emotiv username and password and click on Login. If you do not yet have a cloud username and password, you will need to setup this up on the Emotiv website. A cloud account is not required to use the headset but you will have to go through the training process each time you play Brain Fu unless you have one.

Facial Expressions

To use the Emotiv Insight to answer questions, you must assign actions to the first and second question. The left wink and right wink actions are probably the easiest for the headset to recognize. Click the "First" button on the bottom left of the Emotiv Setup page, and move the selector to Wink Left. Repeat the process on the "Second" page this time setting the action to Wink Right.

There are two options here that will make wink detection more effective. Expression Threshold controls how sensitive the detection logic will be. The higher the threshold the less often an expression will be detected. If a wink is being detected too often (false positive) then raise this threshold until detection stops. If a wink is not detected enough, lower this threshold until the wink is detected more often. Scroll up to the top of the page to see which action is currently being detected.

Repsonse Time is how long it takes to trigger the action before the answer is submitted. The default of 2 should work quite well for facial expressions. The higher this value is, the longer you will have to be winking to answer the questions. You might have to wink a few times before the action is recognized.

The smile and clench facial expressions require a bit of training before they will be detected reliably. Click the Train Smile button to make the training dialog appear. Begin smiling and then press the Start button and hold the smile for 8 to 10 seconds (or until training is complete). To improve the accuracy of the training, do this a few more times. If at any time you want to clear the training data and start over, click Clear Smile.

One you are done training the facial expression, click the Neutral button on the bottom of the page and then click on "Facial Neutral" and then click "Start". Hold a neutral facial expression until training is complete (about 8 seconds). You should train this multiple times to get best results. A lot of neutral training data is important.

Important

You must train at least one session of neutral data otherwise the facial expression will not be deteceted.

Mental Commands

The Emotiv Insight also supports mental commands which means you can think about an action to answer a question. Push, Pull, Lift, Drop, Left, Right and Vanish are the cognitive commands that the headset can recognize. Any one of these can be assigned to an answer so you can play a game of trivia by thinking about pushing or pulling for example.

There are a number of options to fine tune mental command detection. Sens aka Sensititivy allows you to configure how sensitive the detection engine will be for this mental command. The default is 5 which is sometimes too sensitive. Increase this value if you are getting a lot of false positives.

Cache should always be left on. The headset allows you to train the same command multiple times to improve recognition accuracy. The cache must be on to enable this feature. Cache size determines the maximum number of training patterns for single command that will be used by the recognition engine. The default of 20 is a good value.

Activation Level is used once a command is recognized by the headset. It determines how much power a command has. This translates into how quickly Brain Fu will respond to mental commands when they are detected. The default of 5 is a good value. You don't want to set this too high because sometimes you get false positives. Response Time is used in conjunction with Activation Level to determine how long a mental command will have to be registered before it is recognized as an answer. The higher Response Time is, the longer you will need to think about a command to have it recognized.

Important

Be sure to train Mental Neutral many times. To get good quality recognition of mental commands you must do this a lot. The more you can train, the better. The headset needs a good idea of what your mind at rest looks like. You must train mental neutral at least once otherwise the headset will not recognize ANY mental commands at all.

Try not to move and muscles when performing training because this could confuse the headset. Don't make a pushing motion, just picture yourself calmly pushing a box accross the floor. Over exerting yourself could impair command detection.

My Experience

I had great results with the facial expressions. Winks and Smiles are easily recognized with very high accuracy and almost no false positives.

I didn't have much luck with mental commands. I received a lot of false positives and it was hard to sustain the command long enough to answer the question in game. Tuning the settings helps improve the experience a bit. From what I've read it takes a lot of practice using mental commands before you get good accuracy.

My best advice is to train as much Neutral data as you can and try to get the skill level of your training as high as possible. For this you will need to be in a calm environment that allows you to focus.