Heli Koskimäki

Activity recognition using a wrist-worn inertial measurement unit: A case study for industrial assembly lines

As wearable sensors are becoming more common, their utilization in real-world applications is also becoming more attractive. In this study, a single wrist-worn inertial measurement unit was attached to the active wrist of a worker and acceleration and angular speed information was used to decide what activity the worker was performing at certain time intervals. This activity information can then be used for proactive instruction systems or to ensure that all the needed work phases are performed. In this study, the selected activities were basic tasks of hammering, screwing, spanner use and using a power drill for screwing. In addition, a null activity class consisting of other activities (moving around the post, staying still, changing tools) was defined. The performed activity could then be recognized online by using a sliding window method to divide the data into two-second intervals and overlapping two adjacent windows by 1.5 seconds. Thus, the activity was recognized every half second. The method used for the actual recognition was the k nearest neighbor method with a specific distance boundary for classifying completely new events as null data. In addition, the final class was decided by using a majority vote to classifications of three adjacent windows. The results showed that almost 90 percent accuracy can be achieved with this kind of setting; the activity-specific accuracies for hammering, screwing, spanner use, power drilling and null data were 96.4%, 89.7%, 89.5%, 77.6% and 89.0%, respectively. In addition, in a case with completely new null events, use of the specific distance measure improved accuracy from 68.6% to 82.3%.

Gamified physical activation of young men – a Multidisciplinary Population-Based Randomized Controlled Trial (MOPO study)

BackgroundInactive and unhealthy lifestyles are common among adolescent men. The planned intervention examines the effectiveness of an interactive, gamified activation method, based on tailored health information, peer networks and participation, on physical activity, health and wellbeing in young men. We hypothesize that following the intervention the physical activation group will have an improved physical activity, as well as self-determined and measured health compared with the controls.Methods/designConscription-aged men (18 years) attending compulsory annual call-ups for military service in the city of Oulu in Finland (n = 1500) will be randomized to a 6-months intervention (n = 640) or a control group (n = 640) during the fall 2013. A questionnaire on health, health behaviour, diet and wellbeing is administered in the beginning and end of the intervention. In addition, anthropometric measures (height, weight and waist circumference), body composition, grip strength, heart rate variability and aerobic fitness will be measured. The activation group utilizes an online gamified activation method in combination with communal youth services, objective physical activity measurement, social networking, tailored health information and exercise programs according to baseline activity level and the readiness of changes of each individual. Daily physical activity of the participants is monitored in both the activation and control groups. The activation service rewards improvements in physical activity or reductions in sedentary behaviour. The performance and completion of the military service of the participants will also be followed.DiscussionThe study will provide new information of physical activity, health and health behaviour of young men. Furthermore, a novel model including methods for increasing physical activity among young people is developed and its effects tested through an intervention. This unique gamified service for activating young men can provide a translational model for community use. It can also be utilized as such or tailored to other selected populations or age groups.Trial registrationClinicalTrials.gov Identifier: NCT01376986

Application of the Extended

Resistance spot welding is used to join two or more metal objects, and the technique is widely used in, for example, the automotive and electrical industries. This paper introduces the use of the k-nearest-neighbor (knn) method to identify similar welding processes. The two main benefits achieved from knowing the most similar process are the following: 1) The time needed for the setup of a new process can be substantially reduced by restoring the process parameters leading to high-quality joints, and 2) the quality of new welding spots can be predicted and improved using the stored information of a similar process. In this paper, the basic knn method was found to be inadequate, and an extension of the knn method, which is called similarity measure, was developed. The similarity measure provides information of how similar the new process is by using the distance to the knns. Based on the results, processes can be classified, and the similarity measure proved to be a valuable addition to the existing methodology. Furthermore, process information can provide a major benefit to welding industry.

k

A novel method to classify long-term human activities is presented in this study. The method consists of two parts: quick test and periodic classification. The quick test uses temporal information to improve recognition accuracy, while the periodic classification is based on the assumption that recognized activities are long-term. Periodic quick test (PQT) classification was tested using a data set consisting of six long-term sports exercises. The data were collected from six persons wearing a two-dimensional accelerometer on their wrist. The results show that the presented method is not only faster than a normal method, that does not use temporal information and does not assume that activities are long-term, but also more accurate. The results were compared with a normal sliding window technique which divides signal into smaller sequences and classifies each sequence into one of the six classes. The classification accuracy using a normal method was around 84% while using PQT the recognition rate was over 90%. In addition, the number of classified sequences using a normal method was over six times higher than using PQT.

nn Method to Resistance Spot Welding Process Identification and the Benefits of Process Information

This study concentrated on real-time monitoring of a worker using wearable-sensor-based activity recognition. An inertial measurement unit was attached to both wrists of the worker and, by using acceleration and angle speed information, the activities performed by the worker were recognized. Online recognition was done using the sliding window method to divide the data into two-second intervals, and the activity performed in each window was recognized using a knn classifier. Moreover, especially the real-time aspect of the system was considered by studying ways to decrease the amount of training data needed in knn recognition. The studied methods for decreasing the data set included feature and instance selection methods. The results show that instance selection is the best solution for data reduction, while the overall recognition accuracy of the system is high enough for use on industrial assembly lines. When an independent test set is used, the recognition rates using an instance selection method are on average 79.2% which is six percentage units better than the best result using feature selection methods.

Periodic quick test for classifying long-term activities

The activity recognition approaches can be used for entertainment, to give people information about their own behavior, and to monitor and supervise people through their actions. Thus, it is a natural consequence of that fact that the amount of wearable sensors based studies has increased as well, and new applications of activity recognition are being invented in the process. In this study, gym data, including 36 different exercise classes, is used aiming in the future to create automatic activity diaries showing reliably to end users how many sets of given exercise have been performed. The actual recognition is divided into two different steps. In the first step, activity recognition of certain time intervals is performed and in the second step the state-machine approach is used to decide when actual events (sets in gym data) were performed. The results showed that when recognizing different exercise sets from the same occasion (sequential exercise sets), on average, over 96 percent window-wise true positive rate can be achieved, and moreover, all the exercise events can be discovered using the state-machine approach. When using a separate validation test set, the accuracies decreased significantly for some classes, but even in this case, all the different sets were discovered for 26 different classes.

User-independent activity recognition for industrial assembly lines-feature vs. instance selection

Nowadays, huge amounts of information from different industrial processes are stored into databases and companies can improve their production efficiency by mining some new knowledge from this information. However, when these databases becomes too large, it is not efficient to process all the available data with practical data mining applications. As a solution, different approaches for intelligent selection of training data for model fitting have to be developed. In this article, training instances are selected to fit predictive regression models developed for optimization of the steel manufacturing process settings beforehand, and the selection is approached from a clustering point of view. Because basic k-means clustering was found to consume too much time and memory for the purpose, a new algorithm was developed to divide the data coarsely, after which k-means clustering could be performed. The instances were selected using the cluster structure by weighting more the observations from scattered and separated clusters. The study shows that by using this kind of approach to data set selection, the prediction accuracy of the models will get even better. It was noticed that only a quarter of the data, selected with our approach, could be used to achieve results comparable with a reference case, while the procedure can be easily developed for an actual industrial environment.

Recognizing gym exercises using acceleration data from wearable sensors

A database system for storing information on resistance spot welding processes is outlined. Data stored in the database can be used for computationally estimating the quality of spot welding joints and for adaptively setting up new welding processes in order to ensure consistent high quality. This is achieved by storing current and voltage signals in the database, extracting features out of those signals and using the features as training input for classifier algorithms. Together the database and the associated data mining modules form an adaptive system that improves its performance over time. An entity-relationship model of the application domain is presented and then converted into a concrete database design. Software interfaces for accessing the database are described and the utility of the database and the access interfaces as components of a welding quality assurance system is evaluated. A relational database with tables for storing test sets, welds, signals, features and metadata is found suitable for the purpose. The constructed database has served well as a repository for research data and is ready to be transferred to production use at a manufacturing site.

Two-level clustering approach to training data instance selection: A case study for the steel industry

The amount of studies on classification of human characteristics based on measured individual signals has increased rapidly. In wearable sensors based activity recognition a common policy is to report human independent recognition results using leave-one-person-out cross-validation scheme. This can be a suitable solution when feature or model parameter selection is not needed or it is done outside the validation scheme. Unfortunately, this is not always the reality. Thus in this article it is studied how the train-validate-test approach changes the recognition rates compared to basic leave-one-out cross-validation approach. Results of three different ways to perform the train-validate-test is presented: 1) single division to training and testing data, 2) 10-fold division to training and testing data, and 3) double leave-one-person-out cross-validation. In this article, it is shown that the best classifier or feature set selected based on the training and validation data using basic leave-one-out approach does not always perform best within independent testing data. Nevertheless, a larger bias to results can be achieved using single division or even 10-fold division into separate training and testing data. Thus it is stated that the double leave-one-person-out is the most robust version for reporting classification rates in future studies of activity recognition as well as other areas where human signals are used.

Building a Database to Support Intelligent Computational Quality Assurance of Resistance Spot Welding Joints

Many governments and institutions have guidelines for health-enhancing physical activity. Additionally, according to recent studies, the amount of time spent on sitting is a highly important determinant of health and wellbeing. In fact, sedentary lifestyle can lead to many diseases and, what is more, it is even found to be associated with increased mortality. In this study, a data set consisting of self-reported questionnaire, medical diagnoses and fitness tests was studied to detect sedentary young men from a large population and to create a profile of a sedentary person. The data set was collected from 595 young men and contained altogether 678 features. Most of these are answers to multi-choice close-ended questions. More precisely, features were mostly integers with a scale from 1 to 5 or from 1 to 2, and therefore, there was only a little variability in the values of features. In order to detect and profile a sedentary young man, machine learning algorithms were applied to the data set. The performance of five algorithms is compared (quadratic discriminant analysis (QDA), linear discriminant analysis (LDA), C4.5, random forests, and nearest neighbours (kNN)) to find the most accurate algorithm. The results of this study show that when the aim is to detect a sedentary person based on medical records and fitness tests, LDA performs better than the other algorithms, but still the accuracy is not high. In the second part of the study the differences between highly sedentary and non-sedentary young men are searched, recognition can be obtained with high accuracy with each algorithm.