We aim to help adhere and motivate the society towards a healthy tomorrow by empowering the individuals with a personalized health and fitness coach which relies on the power of the smart phones (sensors and computation).
Given to the increasing percentage of unhealthy population in today’s society which can be attributed to sedentary unhealthy lifestyle and excessive food intake, preventive and recuperative healthcare becomes utmost important. Thus, a coaching application, which gives real-time, personalized coaching considering ones physical, mental and emotional state will go a long way. The task becomes somewhat easy with the heightening adoption
and ubiquity of smartphones. For this project the challenge is how effectively we can motivate people towards a healthy and fit regime by giving them personalized feedback taking into account their physical, mental and social wellbeing.
From the science standpoint the project delves into nuances and analysis of human movement by quantifying it from phone sensors’ readings. We then interpret, besides other parameters, the stride frequency, heart rate and speed aimed at efficient and injury free workout. This is also most critical for feedback on quality of workout (and not just quantity). For the efficient utilization of device resources especially the battery, we are implementing a dynamic phone-cloud platform which intelligently manages the communication and computation between phone and the cloud. Since coaching and feedback is one of the primary aims, we utilize the historical personal data for smart analytics, which includes social analytics as well. Finally, the project aims to work on the security of the sensor data on the cloud. The intention is to assure privacy of the users in multi-user scenario and also support fine-grained access control.
The idea of the project came to light by insufficiencies in existing personalized health and fitness solutions. Most exhibit a part (never whole) of what we are trying to achieve in this project. Also, not many of them leverage the computational and sensing power of smart phones, and mostly rely on external hardware.
Biggest results so far
Advising optimal step frequency for runners
Every year about one-fifth of the 2,5 million Dutch amateur runners gets injured. Running with too large steps, and therefore mostly with a frequency that is too low, is known to increase the injury risk of a runner. The optimal step frequency of a runner depends on the heart rate and the running speed, and differs between individuals. Beginning runners are known to have step frequencies below their energetic optimum. Using the smartphone as a sensing tool, a beginning runner can be guided to increase step frequency if necessary. From previous runs an optimal step frequency can be calculated. To this end, we develop a robust step frequency algorithm for unconstrained smartphones and calculate individual optimal step frequencies from training data. More.
ICT science question: how can we use parameters measured of a variety of signals to optimize a desired response? Robust algorithms are needed to work beyond well-controlled environments of a laboratory. And another challenge is the time-variations in the signal.
Involved COMMIT/partners: Infosys, VU Amsterdam, Almende, University of Twente, Sense Labs and HvA.
Run, talk and don’t get injured
We have developed a smartphone app that provides a novel, automated and unobtrusive way of assessing a runner’s physical state based on speech. Beginning runners often have difficulty determining whether they are exercising at the right intensity. They often start with running too fast, and this increases the risk of exercise dropout and injuries. One of the most widely used subjective ways to assess the level of exercise intensity is the ‘talk test’. When you can still speak comfortably while running, you are running at the right intensity. Our smartphone app analyzes your voice and indicates whether or not you are exercising too hard. More.
ICT science question: how does speech production change under influence of various conditions? How can we develop an algorithm that uses this knowledge for reliable automatic voice-based assessment? To solve these challenges, we train classifiers to gauge two signals while they are being acquired each under individually varying circumstances
Involved COMMIT/partners: VU Amsterdam, University of Twente, HvA and Radboud University.
Intelligently grouping amateur runners
Currently, most amateur runners train alone, which may lead to decreased motivation over time. We have developed an app that intelligently groups runners with similar physical parameters in order to make the training more efficient and more fun. In our demo, we display interactively on a screen how a person running on a treadmill follows a track with other runners in a virtual environment. During the run, the app tries to match her with other groups of virtual runners that she passes by, based on the speed and fitness of the person. When a good group is found, the runners in the group are coloured in a certain way, to indicate our runner that she should train together with them. More.
ICT science question: can we make an app that clusters signals in groups that are similar? Can we develop a clustering algorithm that is fast, fault-tolerant and has low battery usage? As a solution, we propose a novel peer-to-peer clustering algorithm. To our knowledge, this algorithm is the fastest of its kind and the attempt to cluster runners in real-time has never been made.
Involved COMMIT/partners: Almende, Infosys, VU Amsterdam, University of Twente, HvA and Sense Labs,