Summary for Judging (250 words):
The pomodoro technique is a time-management technique that provides a short time constraint to study. Another radically-different alternative is the flowtime technique, which involves working on a particular task until the subject feels tired. Both of these techniques undermine the spontaneous and unpredictable natures of fatigue and motivation.
The Muse S collects raw EEG data from five electrodes on the brain, primarily from the frontal cortex, controlling executive functioning, and the temporal cortex, which processes sensory input.
A numpy real-time fast Fourier transform algorithm was applied on each of the raw data streams, decomposing each waveform into its five component frequency bands. The amplitude values of each frequency band were averaged between the five channels. To reduce the effect of short-term artifacts, we took the long term average over 20 seconds to eliminate outliers of instantaneous fluctuations caused by immediate changes in environmental stimuli.
To visualize the data, we used O3D.js to create a 3D spectrogram of raw EEG data, and chart.js to visualize the processed data. Both chart widgets are live, using websockets to communicate.
Thresholds for alertness (based on the beta/alpha ratio) and attention (theta/alpha ratio) were determined using baseline tests based on EEG data which were used to determine whether the user was extremely concentrated, less concentrated, or not concentrated, which ultimately determined whether the program decided to stop the user from working or allow them to continue after the minimum working time. This was directly connected to the timer and user interface.
Please note that the following information is simply for more insight into our project.
Flowmodoro is an original program that enhances the learning and working experience by allowing for longer periods of concentrated work. This program supplies the user with a smarter way of working that improves the historic Pomodoro method, allowing individuals to maintain high concentration at work while automatically tracking their productivity to determine for how long they should work and when they should stop.
The Pomodoro Technique is a time management system that focuses on maximum productivity within a short period of time, followed by a break at thirty minute to one hour intervals. Although the Pomodoro method is marketed as a one-size-fits-all solution to increase productivity during a task, there are two main problems with this method.
The first problem is time constraint. The Pomodoro method requires individuals to limit their working time before their work even starts. This leads to hurried completion or a lack of time management. Both of these are undesirable factors because they can distract from the task at hand.
The second problem is a disruption in flowstate. Flowstate is a mental state in which an individual has reached peak concentration and is deeply absorbed in the task at hand. The current Pomodoro method causes the anticipation of an alarm in a work period, which acts as a consistent interruption that inhibits an individual’s ability to get into a flow state. If the alarm doesn’t take you out of flow state, the break most certainly will.
Taking inspiration from two study techniques called Flowtime and Pomodoro (hence Flow-modoro), our technique takes advantage of the best of both worlds. Users are able to do their work and enter a flow state, with the technology making the decisions for them after an initial period of time to extend their work period until they become distracted and stop working for a short time period.
What it Does:
The Muse S collects raw EEG data from four passive electrodes on the brain. The AF7 and AF8 electrodes on the front of the head collect data from the frontal cortex, which controls executive functioning and decision making, while the TP9 and TP10 electrodes collect data from the temporal cortex, which plays a major role in sensory input, language processing, object recognition, and processing of auditory information. The fifth electrode, in the center of the forehead, is a reference sensor.
First, the raw EEG data was collected from each of the four electrodes on the Muse S headband. A numpy real-time fast Fourier transform algorithm was applied on each of the raw data streams and frequency binning was completed to decompose each waveform into its five component frequency bands. Theta frequencies (4-7 Hz) are associated with cognitive processing, memory, and difficult but mundane tasks. Alpha waves (7-12 Hz) are present when individuals are calm and awake, and are often used to monitor relaxation. Beta frequencies (12-30 Hz) become stronger with the use of bodily movements as well as conscious thought and logical thinking. Our program used two main ratios to determine the alertness, and thus concentration, of the user when doing any given task. The first ratio was theta/alpha, which has been shown to be highly correlated with visual and spatial attention. The second ratio was beta/alpha, which is highly correlated with alertness and concentration.
From there, the amplitude values of each frequency band were averaged between the five channels, creating a long term average. The long term average was also calculated over a period of twenty seconds to eliminate outliers of instantaneous fluctuations caused by small changes in environmental stimuli that did not have a major impact on the overall concentration of the user. To visualize the data, we used O3D.js to create a 3D spectrogram of raw EEG data, and chart.js to visualize the processed data. Both chart widgets are live, using websockets to communicate.
Baseline thresholds for alertness (based on the beta/alpha ratio) and attention (theta/alpha ratio) were then determined using baseline tests based on EEG data collected from a user when resting versus performing activities requiring more mental effort, such as playing a concentration game. For this subject, the threshold values that were collected were used to determine whether the user was extremely concentrated, less concentrated, or not concentrated, which ultimately determined whether the program decided to stop the user from working or allow them to continue after the minimum working time. This was directly connected to the timer and user interface. In the future, these thresholds will be used to initially calibrate the product for each user with a series of simple baseline tests to make the product more personalized, as the thresholds will vary from person to person. Thus, with the concentration thresholds and baseline data, the Flowmodoro program can elongate and optimize the user’s study schedule based on their level of concentration.
On the user end, there are two main adjustments we allowed the user to make. First, the user is able to set a minimum work time. The program will not interrupt the user before this point, even if they are distracted or not very concentrated on their work. This allows a window of time for flowstate to begin. The default minimum work time set by the program is 25 minutes. Second, the user is also able to set a maximum work time. The program will automatically end the work session at this time, regardless of concentration levels. The default maximum time set by the program is 90 minutes. Therefore, the program ends the work session when the user loses focus past the minimum time mark, or at the maximum time mark. In conclusion, the time period of the work session will change depending on the levels of alertness and attention measured from the user, and their relation to the original baseline thresholds.
While making our project, we encountered a few challenges, including issues with git, trouble modding the code that we are incorporating, and general unexpected behavior. During the making of the project, we had 2 major difficulties while working with git. The first of which was a merge conflict that led to updated files and outdated files ending up in the same commit. The other was when one of our group members accidentally deleted all the code in the repository and we had to restore it. We also had trouble modifying the code of the projects we were incorporating into our own. We were using an open source pomodoro timer and modifying it so that it would have the grace time aspects and we had a bit of trouble doing that. We also ran into some issues with the 3d spectrogram where we couldn’t figure out how to change the scale from logarithmic to linear. There were also times where programs would behave weirdly, like when running the same program in the same environment on 2 machines would yield different results, or how there would be unexpected artifacts in our audio.
What we learned:
First, we learned about taking a long term average and using frequency binning to get a more accurate measure of the five EEG frequency bands. The long term average of 20 seconds was helpful because it helped us eliminate substantial outliers of instantaneous fluctuations caused by small changes in environmental stimuli that were happening with data that was averaged every second. After taking this average, the overall data was much more accurate.
Second, we noticed that beta values were consistently higher during concentration periods. From a mentor, we learned that higher cognitive load caused by activities involving decision making and intense concentration also cause out-of-sync neurons to fire, causing higher frequencies in the measured voltage fluctuations. Thus, comparing the specific ratio between the beta band and the other bands gave us a user-independent metric of concentration.
Furthermore, we learned that while the Muse S has a high temporal resolution, in that it is able to measure EEG data in extremely short time intervals, it has low spatial resolution, meaning the specificity to the area of the brain the data is coming from is extremely low. At any given moment, lots of different neurons are firing in the brain, some of which are not related to the part of the brain data should be collected from. However, differentiating between these groups of neurons is beyond the capability of the Muse S. Thus, the best we could do to offset this issue was ensure that the data was averaged after being changed through the Fourier transform into frequency bands to allow for as much accuracy as possible. Relating to such limitations, we recognized that spectral characteristics noted in the 3D spectrogram can be influenced by two main factors. First, the region of the brain under the electrode and the neurons that are firing at any given moment in those regions or neighboring regions, and second, the expected brain activity as a result of the specific task being performed (ex. playing a video game vs taking a reading comprehension test). Given this information, we transformed raw EEG data from each passive electrode and collected specific threshold measures based on the type of activity the subject was performing to allow the program to elongate and optimize the user’s study schedule based on their level of concentration.
Finally, an important component of marketing and product design that should be considered is the idea of making a basic working product before adding additional features when marketing. The importance of having a reliable, tested product before the addition of side products is reflected in the marketing plan.
What’s Next and Commercial Viability:
The entire online platform and application can be accessed by users through a subscription. The initial market includes primarily high school and post secondary students who require immense focus and concentration over long periods of time in their schoolwork. The final market may or may not change depending on whether or not a physical product is created. A potential physical product would include a timer, alarm system, lights, and music which would be sold along with the digital application. Regarding increased growth and expansion, there are three factors that would help Flowmodoro reach a wider targeted audience, including those in corporate or secretarial jobs requiring concentrated work. These factors include overseas sales, integration as manufacturing, material, and shipping costs fall with increased sales, and the use of user data for research purposes with consent, which would help create side products and a gradual branching out of the startup.
We have also considered a SWOT (Strengths, Weaknesses, Opportunities, Threats) analysis in our commercial plan. Strengths include the fact that the market for productivity technology involving the Pomodoro technique is relatively new, that the program is extremely user-friendly, and that only a one-time purchase of the Muse (which has a long lifespan) along with a recurring subscription is needed. Weaknesses are that the Muse itself is expensive and the accessibility in ordering the Muse online is limited, depending on where the buyer lives. Opportunities include that Flowmodoro can target those outside of the target audience - anyone who has a need for concentrated work, not just those who are students, is a potential buyer. There is also a potential for creating side physical products and extending out past the original productivity product. Finally, the main threat to this product is that it is not a necessity and thus people may choose not to buy it for that reason.
Wonder where we got our team name?
The name Astrocyte Feet is derived from the foot-like processes found on astrocyte cells. These processes bind together specifically at the Blood Brain Barrier to prevent large foreign molecules from entering the brain. We found this name particularly fitting because we love feet/because we worked as a team to create this project. We hope you like it but we don't hope you like it more than our team name.