Fischer4You
Fischer4you will study skier’s behavior to find their skill level of skiing, therefore provide them with an appropriate range of equipment and ski skill recommendation. This project is motivated by the fact that every group of skiers need skiing equipment not only based on their general profile but also their technical profile. Fischer4you will try to assess the skill level and performance of skiers in driving different skiing techniques upon a data driven model and therefor is related to fundamental research in the interdisciplinary fields of sport science, computer science, biomechanics and kinesiology. In addition, Fischer4you is not limited to performance analysis, as it goes beyond this by also using the data to recommend users products and suggestion to improve their skills according to their assessment.
Alpine skiing techniques are defined by the Austrian skiing teaching union as the following: Glide, Schuss, Wedging, Snowplough, Drift, Parallel Short Swing, Parallel Long Swing, Carving Long Swing and Carving Short Swing. Each Techniques varies in several regards, like speed or turn rate and they also need to be adapted towards the state of the piste. This makes it hard for unexperienced users to have a consistent behavior in driving different techniques and control speed at each turn. Thus they often lose their control at turns and are not able to manage their speed and body angle to make a good curve, resulting in a lot of severe accidents with other skiers. Moreover, there is a correlation between the behavior at each turn, created curve and its radius, and proper equipment specification. Often users tend to use equipment which doesn’t fit their profile or driving style resulting in both a bad experience at skiing as well as endangering others as they cannot fully control their skis.
Therefore, Fischer4you aims to get qualitative data insights into the process of skiing using multi sensor setup and also test the feasibility of using just a single sensor smartphone system for broad usage. These insights should deliver where best to place the smartphone to get the best out of the collected data while skiing. In the end the goal is to provision a data preprocessing and machine learning pipeline to be executed on data sent by a smartphone app. This pipeline detect patterns and insights from accelerometer, gyroscope, magnetometer and GPS signals, as well as derived data such as altitude, such as the skier’s skill from the generated patterns during skiing, make a skier profile and assist them by a feedback in form of product and ski recommendations and technique guiding.
Goals
Fischer4you goal is to recommend the most suitable ski equipment to each ski driver according to their performance profile and making them aware of their skill level and the level of consistency they have during skiing. Fischer4you will monitor (i) user activities occurring during a skiing day using body worn sensors embedded into everyday smartphones, (ii) user behavior while skiing such as speed control and consistency in driving several different techniques (according to the specification of the Austrian ski teaching union), (ii) user motions such as stability at each turn and acceleration/deceleration after/before each turn, to build a skill assessment model and a skier performance profile. Ultimately, this system, provides users their skiing profile, supports them to access their own driving quality and choose a proper product and to improve their skiing skills. Further the goal is that this system runs on a central server and its assessment can triggered by just using data coming from a mobile phone, which is also used for information users about their skill level and an appropriate product recommendation.
Approach
Fischer4you will build on and advance over state-of-the-start methods for performance and motion analysis, activity identification and skill level recognition relying on machine learning models. Fischer4you will acquire data via IMU sensors and GPS sensors, preprocess the received signals, make a skier profile according to the generated patterns and skiing performance, and derive measures of skill level through a data driven model which compares each user with expert users’ behavior, i.e. users which are teaching skiing instructors. Subsequently, skill level fused with the other general information sources, e.g. demographic information, form input to the recommendation system to get both ski product and improvement recommendations.
Expected and Achieved Results
Fischer4you aims to develop a smartphone application to be used by everyone with any level of skill. This recognition pipeline consists of the following components: (i) a sensor fusion module used to create viable sensor ensembles in an opportunistic manner as a basis for motion analysis and activity recognition, (ii) performance analysis and activity recognition models to know what each skier does and how well he is performing each technique, (iii) a data driven model based on expert users profile to be used as a reference for skill assessment (iv) a recommendation module, able to offer the most appropriate equipment to each user, and (v) a feedback trigger module, which will formulate a ski recommendation and potentially also areas of improvements, e.g. like recommending seldomly driving techniques. This framework is to be developed as distributed application, with a sensing and actuating component, e.g. an app on a smartphone, a reasoning component, e.g. implemented as a machine learning server component for recognition, as part of the demonstrator project 1. For demonstration purposes, the produced application on a smartphone along with the skill assessment framework and recommendation system will provide state of the art user assistance to form a stand-alone, visual support- and guidance system for a targeted personas study and for a bigger audience of skiers in the long run.
