No full-text available
To read the full-text of this research,
you can request a copy directly from the authors.
Automatic detection of alpine ski turns in sensor data
Abstract
We experiment with using sensors and a machine learning algorithm to detect and label turns in alpine skiing. Previous work in this area involves data from more sensors and turns are detected using either a physics-based model or custom signal processing algorithm. We recorded accelerometer and gyroscope data using a single sensor placed on a skier's knee. Left and right turns in the data were labeled for use in machine learner. Although skiing data proved to be difficult to label precisely, a classifier trained on 37 labelled examples correctly label all 13 examples from a different test data set with 2 false positives. This method allows for the use of a single sensor and may be generalizable to other applications.
... A survey of existing work on sensor systems for skis reveals a variety of research objectives. Readings from accelerometers have been used to detect alpine ski turns [4], as well as for analysing an athlete's bio-mechanics of ski jump landings [3]. For understanding forces within the ski, a custom made surfacemounted strain gauge has been proposed for measuring turning direction, angle and deflection [15]. ...
... Sensor systems that localize their measurements to one point within a constrained expanse, like the ones seen on [4] are limited in their ability to obtain insightful information that can come from forces acting at the same time on different points of a given surface. The acquisition of this information can reveal insights that would otherwise remain unseen. ...
Distributed forces on a surface can reveal information about the interaction between an object and its surrounding environment. Sensors can be embedded in sporting equipment to measure the response of the apparatus to forces generated by an athlete. In the case of snow skis, these measurements can be used for purposes of enhancing product development and athlete performance, as well as for analyzing different terrains and snow conditions. This paper describes a piezoelectric system that converts distributed mechanical strains into electrical signals, and records them. This low-cost, low-tech system can be embedded on the top side of most snow ski designs in a minimally invasive manner. Descriptions of the circuit design for the prototype sensor, as well as details about its manufacturing process, are provided. A validation procedure is demonstrated along with sample data presented as a frequency spectrogram. These tests confirmed the proper functioning of the electrical and mechanical design. We conclude that the device could potentially be used to record information that yields insights into performance of materials, user actions, and external conditions about the environment in which the ski is being used.
... The noninvasive monitoring of human activities using mobile and wearable devices is gaining considerable attention in various application domains such as fitness, (1) sports, (2) healthcare, (3) and work performance management (4) owing to the enhanced computational and processing capabilities of these devices. In general, machine learning and deep learning technologies are used to identify an activity label of a particular time period, (5) in which a recognition model is trained in advance using a dataset obtained from a certain number of people. ...
... The recognition performance improves when a larger number of people provide their data because of the increasing degree of heterogeneity. (1)(2)(3)(4)(5)(6)(7)(8)(9)(10) Therefore, a large number of people are required to make the recognition system robust for new users; however, it is quite challenging to build a human activity recognition system from a large amount of data with sufficient heterogeneity. ...
In daily life, people perform activities every moment differently from one another. Thus, it is necessary to develop a robust system that can recognize human activities and cope with their individual differences. In this article, we propose a new method of individualizing a classifier by choosing the most suitable one based on the estimation of compatibility with a set of classifiers, which we call compatibility-based classifier personalization (CbCP). To make CbCP effective and reduce the burden on the user, the number of activities that a user needs to perform to provide data should be as small as possible. We propose two methods of ranking activities that are as effective in estimating the compatibility as using all activities: difference-based and correlation-based approaches. Additionally, we evaluated four methods of handling a case when more than two classifiers have the same level of compatibility, i.e., multi-compatible classifier handling: random choice, average compatibility reference, and ensemble classification with and without weighting. An offline experiment was carried out using two public datasets, i.e., PAMAP2 and DaLiAc, to understand the characteristics of these methods. The results showed that the correlation-based method for activity ranking and the average compatibility reference for multi-compatible classifier handling are the best combination in terms of classification performance, the burden on the user, and computational complexity.
... IMU suits are a popular sensing modality in studies that focus on vibration [16] or turn detection [17]. Other studies investigated skier turn detection algorithms for alpine skiing and utilized a small number of IMUs across various locations on the skier body such as the knee [18] or boot cuff [19,20]. Ref. [6] had a similar sensing setup and used in-field data to classify specific skiing maneuver types. ...
Modern ski design is an inherently time-consuming process that involves an iterative feedback loop comprised of design, manufacturing and in-field qualitative evaluations. Additionally consumers can only rely on qualitative evaluation for selecting the ideal ski, and due to the variation in skier styles and ability levels, consumers can find it to be an inconsistent and expensive experience. We propose supplementing the design and evaluation process with data from in-field prototype testing, using a modular sensor array that can be ported to nearly any ski. This paper discusses a new distributed Inertial Measurement Unit (IMU) suite, including details regarding the design and operation, sensor validation experiments, and outdoor in-field testing results. Data are collected from a set of spatially distributed IMUs located on the upper surface of the ski. We demonstrate that this system and associated post-processing algorithms provide accurate data at a high rate (>700 Hz), enabling the measurement of both structural and rigid ski characteristics, and are robust to repetitive testing in outdoor winter conditions.
... Their analysis suggested that some features have a greater impact on the athlete's performance than others. The system in [15] uses accelerometer and gyroscope data to train a classifier and recognize left and right alpine ski turns. The authors in [16] developed a weight-lifting assistant that uses motion sensors mounted on the athlete's body and provides qualitative feedback on the user's performance. ...
Interactive technology is a rapidly evolving scientific field that includes advanced human-computer interaction technologies and interactive multimodal user interfaces. These interactive platforms usually are utilizing modern techniques such as internet of things (IOT), augmented reality AR, virtual reality VR and use of sensors.
Moreover, there is a developing interdisciplinary scientific field which is trying to integrate modern technology into sports. Computer technology may prove a valuable supportive tool for (a) athletes of all ages to improve their technical skills, and (b) advanced coaching.
The aim of this study is to integrate educational and interactive technology into sports in order to support the training of young bowling athletes, raise their engagement to their sport and boost their performance. Following this direction, this research utilizes a system of sensors to apply step tracking of a young athlete during a throw ball attempt.
The proposed solution is a European funded project (ERASMUS+) supporting the “Europe 2020 strategy” that stresses the need for transforming educational content for instruction and training in a way that engages, motivates, and immerses people to develop personal experiences of constructing their learning experiences.
... Their analysis suggested that some features have a greater impact on the athlete's performance than others. The system in [15] uses accelerometer and gyroscope data to train a classifier and recognize left and right alpine ski turns. The authors in [16] developed a weight-lifting assistant that uses motion sensors mounted on the athlete's body and provides qualitative feedback on the user's performance. ...
Bowling is a target sport that is popular among all age groups with professionals and amateur players. Delivering an accurate and consistent bowling throw into the lane requires the incorporation of motion techniques. Consequently, this research presents a novel IoT Cloud-based system for providing real-time monitoring and coaching services to bowling athletes. The system includes two inertial measurement units (IMUs) sensors for capturing motion data, a mobile application, and a Cloud server for processing the data. First, the quality of each phase of a throw is assessed using a Dynamic Time Warping (DTW)-based algorithm. Second, an on-device-level technique is proposed to identify common bowling errors. Finally, an SVM classification model is employed for assessing the skill level of bowler athletes. We recruited nine right-handed bowlers to perform 50 throws wearing the two sensors and using the proposed system. The results of our experiments suggest that the proposed system can effectively and efficiently assess the quality of the throw, detect common bowling errors, and classify the skill level of the bowler.
... Their analysis suggested that some features have greater impact on athlete's performance than others. The system in [17] uses accelerometer and gyroscope data to train a classifier and recognize left and right alpine ski turns. The authors in [33] developed a weight lifting assistant that uses motion sensors mounted on athlete's body and provides qualitative feedback on the user's performance. ...
Bowling is a target sport that is popular among all age groups with professionals and amateur players. Delivering an accurate and consistent bowling throw into the lane requires the incorporation of motion techniques. Consequently, this research presents a novel IoT-Cloud based system for providing real-time monitoring and coaching services to bowling athletes. The system includes two inertial measurement units (IMUs) sensors for capturing motion data, a mobile application and a cloud server for processing the data. First, the quality of each phase of a throw is assessed using a Dynamic Time Wrapping (DTW) based algorithm. Second, an on device-level technique is proposed to identify common bowling errors. Finally, an SVM classification model is employed for assessing the skill level of bowler athletes. We recruited nine right-handed bowlers to perform 50 throws wearing the two sensors and using the proposed system. The results of our experiments suggest that the proposed system can effectively and efficiently assess the quality of the throw, detect common bowling errors and classify the skill level of the bowler.
... Noninvasive monitoring of human activities using mobile and wearable devices has been gaining considerable attention in various application domains such as fitness and sports [1], healthcare [2], and work performance management [3] due to the improvement of the computational and processing capabilities of these devices. Meanwhile, the use of sensors on animals has been employed for ecological research of various animals such as birds, fish, and mammals using electronic tags [4,5]. ...
The miniaturizations of sensing units, the increase in storage capacity, and the longevity of batteries, as well as the advancement of big-data processing technologies, are making it possible to recognize animal behaviors. This allows researchers to understand animal space use patterns, social interactions, habitats, etc. In this study, we focused on the behavior recognition of Asian black bears (Ursus thibetanus) using a three-axis accelerometer embedded in collars attached to their necks, where approximately 1% of data obtained from four bears over an average of 42 d were used. A machine learning was used to recognize seven bear behaviors, where oversampling and extension of labels to the period adjacent to the labeled period were applied to overcome data imbalance across classes and insufficient data in some classes. Experimental results showed the effectiveness of oversampling and a large difference in individual bears. Effective feature sets vary by experimental conditions. However, a tendency of features calculated from the magnitude of the three axes contributing to classification performance was confirmed.
... Note that only one data logger and one video stream are captured. This approach has been used in data collection for Alpine skiing [4], hiking [3], and rock climbing [3]. Synchronization in this process involves capturing, on video, a red flash emitted by the data logger ten seconds after the data logger is turned on. ...
Obtaining labeled data for activity recognition tasks is a tremendously time consuming, tedious, and labor-intensive task. Often, ground-truth video of the activity is recorded along with sensor-data recorded during the activity. The data must be synchronized with the recorded video to be useful. In this paper, we present and compare two labeling frameworks that each has a different approach to synchronization. Approach A uses time-stamped visual indicators positioned on the data loggers. The approach results in accurate synchronization between video and data but adds more overhead and is not practical when using multiple sensors, subjects, and cameras simultaneously. Also, synchronization needs to be redone for each recording session. Approach B uses Real-Time Clocks (RTC) on the devices for synchronization, which is less accurate but has several advantages: multiple subjects can be recorded on various cameras, it becomes easier to collect more data, and synchronization only needs to be done once across multiple recording sessions. Therefore, it is easier to collect more data which increases the probability of capturing an unusual activity. The best way forward is likely a combination of both approaches.
... A second method was developed with the goal of counting and differentiating right and left turns. Jones et al. (2016) proposed a machine learning approach. They placed a customized sensor consisting on an accelerometer and a gyroscope below the skier's knee. ...
Several methodologies have been proposed to determine turn switches in alpine skiing. A recent study using inertial measurement units (IMU) was able to accurately detect turn switch points in controlled lab conditions. However, this method has yet to be validated during actual skiing in the field. The aim of this study was to further develop and validate this methodology to accurately detect turns in the field, where factors such as slope conditions, velocity, turn length, and turn style can influence the recorded data. A secondary aim was to identify runs. Different turn styles were performed (carving long, short, drifted, and snowplow turns) and the performance of the turn detection algorithm was assessed using the ratio, precision, and recall. Short carved turns showed values of 0.996 and 0.996, carving long 1.007 and 0.993, drifted 0.833 and 1.000 and snowplow 0.538 and 0.839 for ratio and precision, respectively. The results indicated that the improved system was valid and accurate for detecting runs and carved turns. However, for drifted turns, while all the turns detected were real, some real turns were missing. Further development needs to be done to include snowplow skiing.
... To the best of our knowledge, VIHapp is the first attempt to support visually impaired people in mountain activities. Other work has explored the use of wearable sensors to support ski [6], ski jumping [2,1] and snowboarding [5,9,8]. Most of these systems were based on motion sensing. 1 https://www.paralympic.org/alpine-skiing/about ...
Many people who are blind or partially blind continue to enjoy alpine sports such as skiing and snowboarding. Because of their restricted visual perception, visually impaired (VI) skiers are usually accompanied by a ski guide. We present VIHapp, a system consisting of smart ski poles, smart armbands and an instrumented ski to enable communication between VI skiers and ski guides. Guides send commands to VI skiers by pressing a button on their ski poles. VI skiers receive commands in the form of vibrations on their armbands. The smart ski detects turns performed by VI skiers and sends notifications to the guide's smart ski poles.
Human activities can be recognized in sensor data using supervised machine learning algorithms. In this approach, human annotators must annotate events in the sensor data which are used as input to supervised learning algorithms. Annotating events directly in time series graphs of data streams is difficult. Video is often collected and synchronized to the sensor data to aid human annotators in identifying events in the data. Other work in human activity recognition (HAR) minimizes the cost of annotation by using unsupervised or semi-supervised machine learning algorithms or using algorithms that are more tolerant of human annotation errors. Rather than adjusting algorithms, we focus on the performance of the human annotators themselves. Understanding how human annotators perform annotation may lead to annotation interfaces and data collection schemes that better support annotators. We investigate the accuracy and efficiency of human annotators in the context of four HAR tasks when using video, data, or both to annotate events. After a training period, we found that annotators were more efficient when using data alone on three of four tasks and more accurate when marking event types when using video alone on all four tasks. Annotators were more accurate when marking event boundaries using data alone on two tasks and more accurate using video alone on the other two tasks. Our results suggest that data and video collected for annotation of HAR tasks play different roles in the annotation process and these roles may vary with the HAR task.
Inexpensive, low-power sensors and microcontrollers are widely available along with tutorials about how to use them in systems that sense the world around them. Despite this progress, it remains difficult for non-experts to design and implement event recognizers that find events in raw sensor data streams. Such a recognizer might identify specific events, such as gestures, from accelerometer or gyroscope data and be used to build an interactive system. While it is possible to use machine learning to learn event recognizers from labeled examples in sensor data streams, non-experts find it difficult to label events using sensor data alone. We combine sensor data and video recordings of example events to create a better interface for labeling examples. Non-expert users were able to collect video and sensor data and then quickly and accurately label example events using the video and sensor data together. We include 3 example systems based on event recognizers that were trained from examples labeled using this process.
The 2016 Ubicomp Ubiquitous Computing in the Mountains (UbiMount) Workshop was held in Heidelberg, Germany on September 12 and 13. An excursion on the second day of the workshop included hiking on trails and rock climbing on the Riesenstein in the Konigstuhl hill of the Odenwald Mountains adjacent to Heidelberg. We recorded accelerometer data and video of hiking and climbing during the excursion. We collected the data and video in order to better understand how to generate labeled events in accelerometer data for us in machine learning of event recognizers for activities done in the mountains. In this paper, we make that data available to the UbiMount community, explain how it was collected and discuss the data collection experience in the mountains.
Many people enjoy hiking as an outdoor escape from everyday life. However, in our survey of 1002 respondants in the United States, 95% of respondents indicate that they prefer to bring their cell phone when hiking. This potentially creates distractions by reminding hikers of the daily pressures they sought to escape. We believe technology has the ability to enable or enhance outdoor activities, such as hiking, without taking away from the experience. But to begin to realize this vision we must first understand current attitudes towards technology in the outdoors. We present data from a nationwide survey conducted within the United States. We have used this data to identify clusters around attitudes towards hiking and technology use while hiking. We present initial results and analysis of this data as well as direction for future work.
Measuring ski deflection while skiing allows the characterization of essential aspects of the complex interplay between the skier's performance and the interaction of the ski with the snow. This provides an opportunity to optimize skis with regard to skiers’ skills and athletic ability, thus improving skiing performance. To establish an analysis system, we developed, characterized and tested a system capable of measuring ski deflection while skiing. Using competition-type slalom skis, we applied 30 strain gauge sensors - 15 on each ski. The strain gauges sampled data at 65 Hz and the readings were translated into deflection estimates. This was done by validating the system using a specialized bending machine equipped with a laser sensor system to accurately track the changes in ski shape with increasing applied force. We sampled strain gauge data while the bending apparatus deformed the ski. The RMS error of our ski-shape estimates relative to laser-measured data was 11 mm. In on slope tests, the center of mass (CoM) position and speed of a skier were acquired using a highly accurate differential Global Navigation Satellite System (dGNSS) at 50 Hz and a pendulum model. From this data we estimated the CoM turn radii of carved turns. During the ski tests, the bending radii along the ski over time were obtained from strain gauge data and analyzed for each carved turn. We observed smaller radii relative to the path radii at the tip of a ski. We further analyzed the correlation between ski turn radius from deflection measurements and the CoM turn radius. The RMS error between radii calculated from deflection measurements and CoM radii obtained with a high resolution dGNSS was on average 1.26m, with the smallest error being 0.78m. We tested our system on a hard-snow slope and a soft-snow slope. Our system has manageable technological complexity and is potentially suitable as a training tool or for use in ski-fitting for skiers of various skill levels.
In this paper we present results related to achieving finegrained activity recognition for context-aware computing applications. We examine the advantages and challenges of reasoning with globally unique object instances detected by an RFID glove. We present a sequence of increasingly powerful probabilistic graphical models for activity recognition. We show the advantages of adding additional complexity and conclude with a model that can reason tractably about aggregated object instances and gracefully generalizes from object instances to their classes by using abstraction smoothing. We apply these models to data collected from a morning household routine.
In many sports, athletes are spatially separated from their coach while practicing an exercise. This spatial separation makes learning new skills arduous because the coach cannot give instructions or feedback on performance. We present the findings of an in the wild study that demonstrate the potential for teaching sport skills with realtime tactile instructions. We focused on snowboard training. Ten amateurs learned a riding technique with a wearable system that automatically provided tactile instructions during descents. These instructions were in sync with the movements of the snowboard and signaled how to move the body. We found that tactile instructions could help snowboarders to improve their skills. We report insights into the snowboarders' opinion and give recommendations for teaching sport skills with tactile instructions. Our findings help to identify the conditions under which tactile instructions can support athletes in sports training.
Obesity is an increasing problem for modern societies, which implies enormous financial burdens for public health-care systems.
There is growing evidence that a lack of cooking and food preparation skills is a substantial barrier to healthier eating
for a significant proportion of the population. We present the basis for a technological approach to promoting healthier eating
by encouraging people to cook more often. We integrated tri-axial acceleration sensors into kitchen utensils (knifes, scoops,
spoons), which allows us to continuously monitor the activities people perform while acting in the kitchen. A recognition
framework is described, which discriminates ten typical kitchen activities. It is based on a sliding-window procedure that
extracts statistical features for contiguous portions of the sensor data. These frames are fed into a Gaussian mixture density
classifier, which provides recognition hypotheses in real-time. We evaluated the activity recognition system by means of practical
experiments of unconstrained food preparation. The system achieves classification accuracy of ca. 90% for a dataset that covers
20 persons’ cooking activities.
Dan Povey, and others. 1997.The HTK book
- Steve Young
- Gunnar Evermann
- Mark Gales
- Thomas Hain
- Dan Kershaw
- Xunying Liu
- Gareth Moore
- Julian Odell
- Dave Ollason
Matthai Philipose, Fine-Grained Activity Recognition by Aggregating Abstract Object Usage
- J Donald
- Dieter Patterson
- Henry Fox
- Kautz
Dan Povey, and others. 1997. The HTK book
- Steve Young
- Gunnar Evermann
- Mark Gales
- Thomas Hain
- Dan Kershaw
- Xunying Liu
- Gareth Moore
- Julian Odell
- Dave Ollason