Periodic Reporting for period 4 - Infantium 2.0 (Adaptive learning platform for personalized children education crossing cognitive and biometric inputs)
Berichtszeitraum: 2016-11-01 bis 2017-04-30
Infantium is developing a cognitive - affective system based on Neuroscience and Affective Neuroscience and Artificial Intelligence to improve the learning process in the 2.0 society. The main features are:
• Measurement of physiological information monitoring the autonomic nervous system to obtain the emotional state of the user based on physiological information
• Face recognition system, to identify the user and to recognize emotions in learning activities in real time
• machine learning algorithms to classify and analyze the resulting data and give personalized insights
nfantium 2.0 will be able to deliver the first adaptative platform in the world based on emotional states and cognitive states, which will improve the way that the next generations learn, both in the school and at home. Our model is a student-centered model, providing teacher and educators more feedback and information regarding their individual students to improve teaching strategies, with specific impact in those populations with learning needs. Now, with the explosion of ICT, there is an opportunity to move towards personalized, learner-centred and variable models that are effective. Every child is completely different and learns different, so our educational models must be transformed to meet their individual needs.
The objective of the infantium project is to be the Google maps of young brains in order to offer personalized learning to each child in the world. We will make this happen through the analysis of the most important factors, while acquiring new knowledge: 1) How kids learn and 2) What really motivates them while learning. By these means, we intend to be the gold standard of young child education. We want to bring this valuable learning repository to all the developers in order to make their content adaptive, changing the way our kids learn on the digital era.
• Measurement of physiological information monitoring the autonomic nervous system to obtain the emotional state of the user based on physiological information
• Face recognition system, to identify the user and to recognize emotions in learning activities in real time
• machine learning algorithms to classify and analyze the resulting data and give personalized insights
nfantium 2.0 will be able to deliver the first adaptative platform in the world based on emotional states and cognitive states, which will improve the way that the next generations learn, both in the school and at home. Our model is a student-centered model, providing teacher and educators more feedback and information regarding their individual students to improve teaching strategies, with specific impact in those populations with learning needs. Now, with the explosion of ICT, there is an opportunity to move towards personalized, learner-centred and variable models that are effective. Every child is completely different and learns different, so our educational models must be transformed to meet their individual needs.
The objective of the infantium project is to be the Google maps of young brains in order to offer personalized learning to each child in the world. We will make this happen through the analysis of the most important factors, while acquiring new knowledge: 1) How kids learn and 2) What really motivates them while learning. By these means, we intend to be the gold standard of young child education. We want to bring this valuable learning repository to all the developers in order to make their content adaptive, changing the way our kids learn on the digital era.
During the period, the second release has been done, with better mechanical design, and better electronics (see photo). This has been a challenging -but really rewarding -period. First, after the great challenges posed by the former subcontractor our in-house hardware team has fixed not only the many problems of the firmware and electronics, but also to greatly improve quality of the signal and pre-processing, and thanks to this, improve the machine learning eliciting of emotional states. We have the first models to use extracted features for classification and regression to infer emotional states.
During the first pre-pilot, as expected, the previous subcontractor work was poor and inefficient and it didn’t allow good contact with the skin, provoking the loss of data. Also we were able to validate our affective system for computer vision using convolutional neural network and external human ratings. The results of these methods gave a 100% concordance between algorithms detected and human ratings.
In fact, computer computer vision software for emotional recognition has improved considerably. Currently, we’re outperforming state of the art in basic emotion analysis, and we’ve been able to get the first classifiers in the world for secondary emotions, specifically for boredom (something that to our knowledge no other company or lab has). in progress, we now implement convolutional neural networks (“deep learning”) and we are inferring a greater set of features (yawns, blink rate -what is related with dopaminergic systems in the body, hand occlusions, eye position)
Regarding exploitation and dissemination of results, we've been actively working on social media, universities and schools to get participants for affective detection and we have uploaded a webapp so we can collect continuously data for our datasets. As planned, we have continued with PR and training work with school, and also we’re now closing a partnership with one of the biggest toy distributors of the country (+200MM in turnover). Also, we’ve been able to close 2 trials with 2 of the best medical institutions of the US for enhancing european-american research and commercialization: Mass General Hospital and McLean Hospital. The trial plans plan to test Robbie system for for monitoring affective states in individuals with Down syndrome, during individuals’ normal activities. The goal of these research projects is to publish in peer review reference papers and journals with the first ever verification of this technology by accredited third parties.
During the first pre-pilot, as expected, the previous subcontractor work was poor and inefficient and it didn’t allow good contact with the skin, provoking the loss of data. Also we were able to validate our affective system for computer vision using convolutional neural network and external human ratings. The results of these methods gave a 100% concordance between algorithms detected and human ratings.
In fact, computer computer vision software for emotional recognition has improved considerably. Currently, we’re outperforming state of the art in basic emotion analysis, and we’ve been able to get the first classifiers in the world for secondary emotions, specifically for boredom (something that to our knowledge no other company or lab has). in progress, we now implement convolutional neural networks (“deep learning”) and we are inferring a greater set of features (yawns, blink rate -what is related with dopaminergic systems in the body, hand occlusions, eye position)
Regarding exploitation and dissemination of results, we've been actively working on social media, universities and schools to get participants for affective detection and we have uploaded a webapp so we can collect continuously data for our datasets. As planned, we have continued with PR and training work with school, and also we’re now closing a partnership with one of the biggest toy distributors of the country (+200MM in turnover). Also, we’ve been able to close 2 trials with 2 of the best medical institutions of the US for enhancing european-american research and commercialization: Mass General Hospital and McLean Hospital. The trial plans plan to test Robbie system for for monitoring affective states in individuals with Down syndrome, during individuals’ normal activities. The goal of these research projects is to publish in peer review reference papers and journals with the first ever verification of this technology by accredited third parties.
The goal is to have a system that is able to reshape adaptive learning environments with maximum impact in young children. Personalized learning is much more than technology evaluating what’s bad or wrong about a student performance. It means understanding how we all learn as humans, understanding that brains are not a motherboard with some engines placed in the same way. Adaptive learning can give equality of opportunities for all. The approach towards adaptive learning must be on Neuroscience, Experimental psychology and learning theory, together with AI like computer vision, affective neuroscience and neural networks.
With this system there’s a better communication between educators-students and families. That is, tracking emotions while learning, cognitive performance when interacting, to create personalized learning. This is to optimize motivation of students, including those with learning disabilities, identifying those students at risk and excelling (individually or groups). This will help teachers concentrate on what’s really important: teaching in smaller groups, and guiding their students to succeed. What’s is necessary to achieve the goal of the EU of reducing early school leavers in the future, and increase skills for growth and jobs in the future, starting from the youngest generation. This is even more important in individuals with challenging behavioral alterations and lack of emotional regulation like special needs (Down Syndrome, ADHD, autism, etc.): 1 billions of people have some kind of special needs representing a financial challenge for authorities and families. In caregiving, greater awareness and attunement to the internal state of an individual can lead to greater wellbeing of both the individuals and their families, overall, 1 billion people that could benefit from it.
Driven by the trends and facts mentioned above, the smart learning market is forecasted to reach $220bn23 by 2017 growing at 20.3 % from 2012 to 2017. It is clear that technology continues to disrupt entire industries, and INFANTIUM wants to be at the forefront of this shift. Currently, there is multiple educators and analysts that considers that the benefits of wearables like the one proposed by INFANTIUM will help classrooms to be more productive and efficient. In fact, a report25 by the NMC publications and cofounded by the EC, states that in four years wearables will foster on educational systems representing the same boom as happening currently on the health or fitness sector. The explosion of the wearable devices opens a wide new world for smart learning applications. Without any doubt, wearable sensors Hardware for emotion sensing at learning process together with interactive software will play a significant role to make the education more effective and pleasant. At Infantium, we want take advantage of this increasing trend and be the first company worldwide offering a wearable device to understand what are the motivations and reactions of each kid while learning.
With this system there’s a better communication between educators-students and families. That is, tracking emotions while learning, cognitive performance when interacting, to create personalized learning. This is to optimize motivation of students, including those with learning disabilities, identifying those students at risk and excelling (individually or groups). This will help teachers concentrate on what’s really important: teaching in smaller groups, and guiding their students to succeed. What’s is necessary to achieve the goal of the EU of reducing early school leavers in the future, and increase skills for growth and jobs in the future, starting from the youngest generation. This is even more important in individuals with challenging behavioral alterations and lack of emotional regulation like special needs (Down Syndrome, ADHD, autism, etc.): 1 billions of people have some kind of special needs representing a financial challenge for authorities and families. In caregiving, greater awareness and attunement to the internal state of an individual can lead to greater wellbeing of both the individuals and their families, overall, 1 billion people that could benefit from it.
Driven by the trends and facts mentioned above, the smart learning market is forecasted to reach $220bn23 by 2017 growing at 20.3 % from 2012 to 2017. It is clear that technology continues to disrupt entire industries, and INFANTIUM wants to be at the forefront of this shift. Currently, there is multiple educators and analysts that considers that the benefits of wearables like the one proposed by INFANTIUM will help classrooms to be more productive and efficient. In fact, a report25 by the NMC publications and cofounded by the EC, states that in four years wearables will foster on educational systems representing the same boom as happening currently on the health or fitness sector. The explosion of the wearable devices opens a wide new world for smart learning applications. Without any doubt, wearable sensors Hardware for emotion sensing at learning process together with interactive software will play a significant role to make the education more effective and pleasant. At Infantium, we want take advantage of this increasing trend and be the first company worldwide offering a wearable device to understand what are the motivations and reactions of each kid while learning.