WO2013083278A2

WO2013083278A2 - Mobile computing system

Info

Publication number: WO2013083278A2
Application number: PCT/EP2012/005052
Authority: WO
Inventors: Bernd TESSENDORF; Franz GRAVENHORST
Original assignee: Eth Zurich
Priority date: 2011-12-07
Filing date: 2012-12-07
Publication date: 2013-06-13
Also published as: WO2013083278A3

Abstract

The present invention relates to a mobile computing system, preferably with at least part of the system being wearable by a user or attachable to a sport or rehabilitation equipment, comprising a sensor system, characterized in that the sensor system contains at least one gyroscope or equivalent sensor to capture rotary orientation and/or motion of the user or of the Sport or rehabilitation equipment. Further, the present invention relates to a method to operate such a mobile computing system.

Description

TITLE

MOBILE COMPUTING SYSTEM

TECHNICAL FIELD

The present invention relates to a mobile computing system, preferably with at least part of the mobile computing system being wearable by a user or attachable to a sport or rehabilitation equipment, the mobile computing system comprising a sensor system. The present invention relates further to a method for operating the same.

PRIOR ART

In the sport of rowing, athletes and coaches are concerned with optimizing a rower's technique in order to improve rowing performance. On-board computers on the boat can provide the rower with the current stroke rate and split times related to speed over the distance covered. However, they don't give any additional feedback on the parameters related to the rower's technique. Visual feedback can be provided by video-goggles [5]. These are worn on the rower's head and display the video signal as it is captured by an accompanying coach. In [6] the authors investigate sonification of boat acceleration as feedback to support synchronization in crew boats. Table 2 gives an overview of sensing approaches to collect data from rowing boats. Several approaches [7] have been suggested to perform biomechanical analysis of rowing technique. Various research has considered indoor rowing, i.e. ergometers [8-10] or rowing simulators [1 1]. The following Table 2 gives an overview of sensing approaches in rowing boats.

Sensor Location Description

Magnet Seat Reed switch to count strokes

Spirit level Boat Support setup of boat stability Impeller Boat Measure speed relative to the water [12]

GPS Boat Measure distance and speed relative to the shore [ 12]

Accelerometer Boat Measure stroke rate [12]

Gyroscope Oars Measure influence on boat stability [13]

Boat Measure boat stability [14,15]

Potentiometer Boat, Oars Measures horizontal oar angle [3]

Force sensor Oarlocks Measure force applied to oars [16]

Foot stretcher Measure force on foot stretcher

Strain gauge Oars Measure bending force on oars

Actual rowing on the water complicates the rowing technique as the boat is not in a fixed position relative to the ground anymore. The boat constantly changes its relative position to the ground, in particular if it does turns or floats on a wavy water surface e.g. during windy weather conditions.

SUMMARY OF THE INVENTION

It is therefore an object of the present invention to further improve the feedback on the parameters related to the rower's technique for trainings on ergometers and, in particular for rowing on water.

This and other objects are achieved by a mobile computing system, preferably with at least part of the system being wearable by a user or attachable to a sport or rehabilitation equipment, comprising a sensor system, wherein the sensor system contains at least one gyroscope or equivalent sensor to capture rotary orientation and/or motion of the user or of the Sport or rehabilitation equipment.

Preferably, the sensor system comprises at least two sensors, wherein the mobile computing system is adapted to calculate a differential signal between at least two of the sensors of the sensor system. It is advantageous to calculate said differential signal as it contains information about the relative movement between the at least two sensors from which the differential signal is originating. Preferably, the mobile computing system comprises a sensor system that includes as sensors at least one gyroscope and at least one additional inertial measurement sensor, wherein at least one sensor is fixed on a boat and at least one sensor, preferably a gyroscope, is fixed on an oar, and wherein the mobile computing system is adapted to calculate a differential signal between said sensors. Here, an inertial measurement sensor is a sensor like an acceleration sensor, magnetic field sensor or a gyroscope that measures in at least one dimension, preferably in two or three dimensions or it is a complete inertial measurement unit which preferably includes triaxial acceleration and magnetic field sensors and a triaxial gyroscope.

According to a preferred embodiment, the sensor system comprises two inertial measurement sensors, wherein the inertial measurement sensors are positioned in a boat and on an oar. Typically, the oar is moved relative to the boat; hence it is advantageous to mount a sensor on the oar and another sensor on the boat. The sensor on the oar provides orientation data of the oar and the sensor on the boat provides data of the boat's orientation, wherein the mobile computing system is adapted to calculate a differential signal to extract oar orientation data that is independent of an absolute boat movement. Thereby, useful information on the relative position of the oar with respect to the boat is gained.

According to another preferred embodiment, the mobile computing system as described above, is characterized in that, as sensor system, two inertial measurement units are positioned on a boat and on an oar, wherein the sensor on the oar provides orientation data and the sensor on the boat provides data of the boat's acceleration, and wherein the mobile computing system is adapted to calculate a differential signal from signals of the inertial measurement systems to extract oar orientation data that is independent of an absolute boat movement.

According to another preferred embodiment, at least one sensor is fixed on the sliding seat and at least one sensor is fixed on the boat, wherein the mobile computing system is adapted to calculate a differential signal to extract sliding seat position data that is independent of an absolute boat movement. Thereby, useful information on the relative position of the sliding seat with respect to the boat is gained.

Preferably, the mobile computing system is adapted to convert sensor values of different sensors into one coordinate system for the purpose of compensating for a spatial misalignment between the individual sensors. Preferably, the coordinate system is a coordinate system of one of the sensors of the sensor system.

Preferably, the mobile computing system comprises further sensors for motion and magnetic field detection, preferably a complete inertial measurement unit (IMU), capturing acceleration, magnetic fields and rotation.

Preferably, the mobile computing system comprises a programmable memory unit, wherein the programmable memory unit can be programmed with customized software for a specific sport or rehabilitation activity, in particular for rowing, so that meaningful information for the specific sport or rehabilitation activity is rendered based on captured sensor data and made available to a recording and/or a user feedback unit, preferably a display unit, a sonification unit, a vibrational unit, or the like.

Preferably, the mobile computing system comprises further a GPS receiver.

Furthermore, the mobile computing system is preferably characterized in that a part of the system, preferably a part comprising a user feedback and/or recording unit, is separated from the sensor system and linked to it via a communications channel, preferably a wireless communications channel.

Preferably, at least the part of the system containing the sensor system is attached to or incorporated in a wrist-worn or feet-worn device, preferably a watch.

Preferably, at least the part of the system containing the sensor system is attached to or incorporated in a piece of clothes, preferably a shirt, cap or shoe.

Preferably, at least the part of the system containing the sensor system is attached to or incorporated in a piece of sports or rehabilitation equipment, preferably an oar or paddle. The above mentioned object is also achieved by a method to operate a mobile computing system as described above, whereas part of the mobile computing system that comprises a sensor system is attached to the user's body or a moveable object or flexible object, preferably to an oar, paddle or to the wrist of a user, and whereas the sensor data is interpreted towards meaningful information concerning the performed activity or movement, preferably for rowing or paddling sports.

According to a preferred embodiment of the method, a differential signal is calculated between at least two sensors of the sensor system and preferably interpreted as described above.

Preferably, the sensor values are converted into one coordinate system in order to compensate for a spatial misalignment between the individual sensors, wherein, preferably, said coordinate system is a coordinate system of one of the sensors of the sensor system. Preferably, meaningful information is calculated from the sensor data, in particular from the differential signal, taking into consideration the rotary orientation information of the oar, paddle or wrist.

Meaningful information is considered to be taken to the following group of data comprising: a. the trajectory of the horizontal and/or vertical and/or rotation oar/paddle angle b. the timing and angular position of the stroke's start (catch),

c. the timing and angular position of the stroke's end (finish),

d. the stroke length,

e. the stroke rate,

f. the classification of the stroke phase (drive or recovery),

g. the ratio drive versus recovery,

h. the timing and angular position of the blade's squaring or feathering movement, 1. the timing and angular position of the blade when it dips into the water or leaves the water.

Preferably, the method as described above involves a step that evaluates the meaningful information against a quality of the user's movement and/or suggests changes in his movement.

Furthermore, the present invention relates to a method to operate a plurality of mobile computing systems, each according to an embodiment as described above, whereas each of the mobile computing systems comprises a storage and/or communication unit and data from all mobile computing systems is collected and arranged in a synchronized way, which means that the sensor data from one individual mobile computing system of the plurality of mobile computing systems can be compared with the sensor data which has been acquired at the same time from another mobile computing system of that plurality of mobile computing systems.

According to a preferred embodiment each of the mobile computing systems comprises additionally a GPS receiver and the method to operate a plurality of mobile computing systems is characterized in that the captured data from the sensor system is linked to a received GPS system time before being stored and/or transmitted.

Here, preferably, each of the mobile computing systems is attached to the wrist, oar or paddle of a team member of a rowing or paddling team, wherein the individual mobile computing systems are operated according to a method as described above, and wherein the collected meaningful information is evaluated towards synchrony between team members.

BRIEF DESCRIPTION OF THE DRAWINGS

Preferred embodiments of the invention are described in the following with reference to the drawings, which are for the purpose of illustrating the present preferred embodiments of the invention and not for the purpose of limiting the same. In the drawings,

Fig. 1 shows the iterative process to optimize a rower's technique: The standard approach and extension using a sensor network approach;

Fig. 2 shows one cycle of rowing [1]: catch phase (1,2), drive phase (3,4), finish

(5) and recovery phase (6);

Fig. 3 shows definitions of the three oar orientation angles: rotation (a), horizontal

(b), and vertical (c);

Fig. 4 shows typical signals for the three orientation angles measured on an oar with our system; shown are 3 strokes at a stroke rate of 20 strokes per minute;

Fig. 5 shows an implementation of the sensor network: three IMUs on the oars and the boat, indicated by red arrows;

Fig. 6 shows results of a continuous monitoring of the stroke rate, stroke length, ratio recovery/drive and ratio feathered/squared for both a typical racing and training session;

Fig. 7 shows a visualization of data of stroke length and stroke rate from two world-class rowers (Rl and R2) for training and racing sessions; Fig. 8 shows a stroke length in detail: In this example the variability of the stroke length results mainly from a variance in the finish position;

Fig. 9 shows an oar rotation angle over the horizontal oar angle for rowers Rl and

R2 for both training and racing session; for visualization the curve for R2 is shifted 10° to the lower left corner; and

Fig. 10 shows a boat acceleration in driving direction over horizontal oar angle to illustrate how the movement of the oar through the water accelerates the boat.

DESCRIPTION OF PREFERRED EMBODIMENTS

In the sport of rowing, athletes and coaches are concerned with optimizing a rower's technique in order to improve rowing performance. Here, we present the design and real- world evaluation of a sensor network approach to support improving the rower's performance. In cooperation with professional rowing teams, we found that a network of inertial measurement units (DVIUs) is well suited to continuously and unobtrusively monitor important indicators relating to rowing technique. In a feasibility study with 5 participants we first investigated the optimal sensor setup, and in the final setup we attached 3 DVIUs to the oars and the boat. From 18 participants (including both ambitious amateurs and world-class rowers) we recorded both training and racing sessions which each consisted of 1000m rowing. We present 4 rowing technique indicators for all 18 participants. Using the example of two world-class rowers we demonstrate in detail how sensor networks support the iterative process of optimizing the individual rowing technique.

I. Introduction

The objective in a rowing race is to move the boat as fast as possible from the start to the finish. Besides their physical and mental strength, the rower's technique is a key feature leading to success. Here, we investigate how to beneficially integrate inertial measurement units (DVIUs) into the iterative process of optimizing rowing technique that is illustrated in the upper part of Fig. 1. We address the following research questions: (1) Measurability: Can we use EMU data to quantify and visualize important rowing technique indicators such as stroke length and rhythm? (2) Accuracy: Is the system accurate enough to analyze the rowing technique and provide benefits for the optimization of rowing technique? (3) Practicality: Can we devise a robust system for daily use in "real life" which is sufficiently unobtrusive to not impair the rower when rowing?

In section II we describe the process of improving rowing technique and a novel approach to support this process. In section III we present the implementation of this approach using a sensor network and dedicated algorithms. In section IV we describe the results for our proposed system in a real-world application and conclude in section V. II. Theory

A. Rowing Technique Basics

The basic rowing technique is illustrated in Fig. 2. Rowing is a periodic movement. One cycle of rowing is explained in the following. The numbers given in brackets refer to the pictures in Fig. 2. During the catch phase (1, 2) the oar blades are placed in the water. The catch is a quick accurate hand movement and should happen as early as possible to maximize the stroke length [2]. In crew boats (more than one rower in a boat) it is essential that all rowers perform the catch at the same time. In the drive phase (3) the legs are extended and then the body opens up (4), levering the boat forwards. The arms are straight and the lower back is firm. In the finish (5) the oars come out of the water and are feathered parallel to the water surface to minimize air resistance. In the recovery phase (6) the rower's limbs move towards the stern in preparation for the next stroke: first the arms, then the upper body and finally the legs.

B. Improving Rowing Technique

Performance in rowing is measured with a recorded time from start to finish of a race. There are qualitative rowing technique guidelines from national and international rowing associations to explain why the time was achieved and how to improve it [2-4]. Table 1 lists a subset of these pieces of best-practice advice that allow further in-depth analysis of rowing technique. Key aspects include stroke efficiency, rhythm, timing, and boat stability. There are quantitative (e.g. velocity, stroke count) and qualitative (e.g. smoothness of movement) measures, but there is a strong need for both rowers and coaches to get more quantitative measures.

The following Table 1 shows a subset of rowing technique indicators advised by international and national rowing associations [2]-[4].

Rowing Technique Indicator Description

Consisted pattern and length Each stroke should resemble the previous one (constant timing). The rower should stick to a long stroke length, even with high stroke rates like during racing conditions.

Good blade depth To avoid losses but to maximize acceleration, the blade should not be too deep or too little in the water during the drive phase. Ideally, the deepness is equal to the blade's height, so the upper edge is just covered with water. During the recovery phase there should be a gap between blade and water approximately as high as the blades' width.

Firm, direct and consistent The blade should go into the water immediately after the action of the blade most forward position (catch position) is reached. At the end of the stroke, the hands and the upper body should start moving towards the catch position without delay. The squaring of the blade should be early enough to be prepared for the catch.

Relaxed, but controlled body The horizontal moving velocity of the blade during the movement during the recovery recovery phase should be constant without impulsive accelerations.

Powerful, smooth body Smooth movement of the boat, jerky peaks in the movement acceleration curve should be avoided.

Synchronicity in crew boats All members of a crew should perform the movements with exactly identical timing.

According to the FISA (The Federation Internationale des Societes d'Aviron (FISA) is the international rowing federation) the long-term objective for the athlete is to master the rowing technique [2]. The goal is to maximize propulsion in the drive phase and to minimize the loss of speed due to friction during the recovery phase. To improve the rower's technique the coach needs to identify the necessary changes in the movement. This is a challenging task because even small changes in rowing technique can significantly affect the finish of a race. Moreover, any fine-tuning and beneficial changes to rowing technique are specific to the particular rower's anatomy and rowing style and therefore have to be found for each rower individually. An analysis of Olympic rowing champions confirms that many different kinds of technique lead to success [4]. In our discussion with professional coaches we found that an iterative trial and error approach is used to optimize the rower's technique as illustrated in the upper part of Fig. 1. To assess the performance and to identify improvement opportunities the coach observes the rower's technique and also takes into account feedback from the athlete. The rower then attempts to make the changes in rowing technique suggested by the coach and the next iteration starts. This approach's success depends on the experience of the coach, who needs to identify shortcomings and suggest appropriate steps based on their own personal qualitative observations. Therefore, novel approaches, which provide additional information to support the coach in this process are highly appreciated.

C. IMU-based Support to Optimize Rowing Technique

To address the need to provide the coach with additional information for optimizing a rower's technique, we have investigated the application of sensor networks in rowing boats. These sensors enable quantitative measurements of certain aspects of rowing technique which are related to the overall rowing performance desired. Our approach is to continuously monitor oar and boat orientation for each rowing stroke and subsequently quantify and visualize a subset of established rowing technique indicators. The lower part of Fig. 1 illustrates the proposed extension of the standard approach. The length of each rowing stroke is just one example of a measure, which coaches cannot observe accurately just with their eyes. Sensor networks can support the coach in this aspect.

We have found that IMUs are well suited to continuously and unobtrusively monitor important indicators relating to rowing technique such as the stroke length and the stroke rate. Fig. 4 depicts typical signals for the three orientation angles measured with an IMU on the oar. The definition of the orientation angles is given in Fig. 3.

As shown in Fig. 4 we calculate the catch and finish phase from the horizontal oar angle. Based on this we deduce the stroke rate, the stroke length, the ratio of recovery/drive, and the variance of these parameters. Based on the oar rotation angle we determine for each point in time if the blade is feathered or squared. The vertical oar angle indicates the depth of the blade relative to the boat. In the following we will focus on IMUs as they represent promising sensors for our task. III. Implementation of an HVIU-based Sensor Network for Rowing Boats A. Pre-study

The design of a sensor network for rowing boats in naturalistic, harsh environments requires that the device is robust. Moreover, when deploying the system in real-world settings it is necessary that the rower is not hindered by a bulky measurement setup. Therefore, an unobtrusive sensing approach is essential to obtain realistic data.

In a pre-study [17] we optimized the type of sensors, the sensor positions, the number of sensors and the sampling rate of the sensors to achieve a reasonable trade-off between required accuracy and used resources. We visualized data from 12 different sensor locations of 10 recording sessions with 5 participants. We found that two EVIUs, positioned on an oar and on the boat, are sufficient to address our research questions. The sensor on the oar provides the orientation data explained above. The sensor on the boat provides data of the boat's acceleration and allows us to calculate a differential signal to assure that the oar orientation data is independent of the absolute boat movement, e.g. that induced by wavy water or a change of the heading. Movements in rowing happen quickly, especially in the catch phase. For our analysis the system's maximum sampling rate of 60 Hz was sufficient.

To get measurements of a moveable object relative to another moveable object we developed a differential measurement technique.

In case of rowing, the movement of the oar relatively to the boat is of interest and is the basis to derive further meaningful data such as the stroke length. However, the inertial sensors (such as accelerometers or gyroscopes) attached to the oar measure data which describes the translational or rotary movement of the sensor relatively to the ground, not relatively to the boat.

We attach an additional sensor unit to the boat shell. This sensor measures the boat's translational and/or rotary movement relatively to the earth surface.

To derive the oar position relatively to the boat, a first approach is to calculate the difference between the measurement vectors of the two sensors. However, to achieve more accurate measurements, we have to take into account that these two sensors are usually not perfectly aligned to each other and thus their coordinate systems are different. If the misalignment is known, the measurements can be corrected. There are basically two approaches of correcting for these misalignments. This is done by converting all sensor values into the same coordinate system. This common coordinate system can be identical to one of the sensor's inertial coordinate system or another coordinate system such as the earth-fixed coordinate system.

One approach is to take the coordinate system of one of the sensors as reference and convert the values of the other sensor to the same coordinate system (Case 1). Another approach is to convert all values of both sensor nodes to another coordinate system (Case 2).

Case 1 :

One of the sensor's inertial coordinate system is used as common coordinate system:

In this approach, only the relative misalignment of one sensor node to the other has to be known and only one of the sensor's measurement vector's has to be corrected. One possibility to compensate the misalignment is to multiply the measurement vector of one sensor with a rotation matrix [18], which transforms the measurements into the coordinate system of the other sensor node: rel_values(n) = rot($ θ, ψ)^■ values _{oar smsor} (ri - values_ioatjgnsofl (n) ( 1 )

This rotation matrix ro£(0, θ, ψ depends on three Euler angles φ, Θ and ψ, which describe the misalignment between the coordinate systems of the two sensor nodes.

There are different alternative ways to obtain the values of these angles:

(a) Static Calibration Procedure

We perform a static calibration procedure each time one of the sensors is relocated. One possible static calibration procedure is to ask the rower to move to a predefined position, for example to move the oar to a position in which all coordinate axes of the oar sensor should be in parallel to those of the boat sensor. In this position we adjust the rotation matrix until the components of the relative measurement vector obtained with equation (1 ) are close to zero.

(b) Exploitation of prior knowledge (movement restrictions)

This method can also be described as "dynamic calibration procedure".

It is similar to method (a), but instead of a static position the rower is asked to perform a predefined dynamic movement. The rotation matrix is then changed accordingly until the output data reflects the expected data of the movement.

For example the boat is held fixed at the dock and the oar handle is moved in a strict horizontal plane from the bow to the stern. The aligned gyro data should only show rotary movements in one axis which corresponds to horizontal rotary movements, gyro values of other axes should be close to zero.

The predefined dynamic movement could also be identical with the normal rowing stroke. This has two advantages:

(i) The rower can immediately begin rowing without having performed unnatural calibration tasks.

(ii) The calibration data can be renewed constantly or periodically during the rowing performance. This enables correction for time-variant differences between both sensors.

(c) A combination of (a) and (b)

Once the initial misalignment is obtained, we have to take into account that during the rowing motion the two sensor nodes and thus their built-in inertial coordinate axes are constantly moving relatively to each other. Depending on the type of data we want to analyze, we have to adjust the rotation matrix constantly. However, this is not necessary for the stroke length. The stroke length is the rotational displacement in the horizontal plane and it is calculated based of the gyro data of the according axis. Due to the restrictions in rowing movements, the position of this axis will not change relatively to the boat-inertial axis.

Case 2:

Another coordinate system is used as common coordinate system:

In this approach a third coordinate system which is usually different from each of the sensor's inertial measurement systems is used as a common coordinate system.

So, the relative misalignment of each sensor node relatively to the common coordinate system has to be known and each one has to be converted separately. This conversion can be done by multiplying the measurement vector with a rotation matrix (see above).

A similar approach is applied for the tracking of the position of the sliding seat relatively to the boat. For this purpose, a differential measurement such as acceleration of the sliding seat relatively to the boat is of interest. This can be obtained by a differential measurement between a sensor node at the sliding seat and one at the boat shell.

B. Sensor Network

In the final setup we used three complete IMUs, each with a triaxial gyroscope, acceleration and magnetic field sensor. As most of the athletes follow a strict training plan we have been granted only a limited amount of time to perform successful recordings. So, we extended the optimal setup found in the pre-study (one sensor at the boat and one at either the starboard or port side oar) with an additional sensor on the second oar for detecting potential asymmetries in the sculling technique and for redundancy. Fig. 5 shows the sensors attached to the two oars and the boat. Each EvlU measures magnet field, acceleration and angle velocity in all three axis to calculate orientation angles. All IMUs are connected to a master device to assure synchronized data. In this study we used a wired system, which stores the recorded data locally. Data recording and synchronization was handled using the Context Recognition Network (CRN) Toolbox [18]. Later, it was transferred to the coach's computer to perform the analysis. The runtime of one set of batteries was about two hours. We experienced no problems concerning data loss. The system can be installed in any type of rowing boat, and is useful for both sculling and sweep rowing. It provides useful data for any size of rowing boat from singles to eights.

C. Algorithms

Based on the raw data from the IMUs we have quantified various rowing technique indicators:

Stroke Rate: The stroke rate is a standard rowing parameter defined as the number of strokes per minute (spm). We compute the stroke rate by counting the frequency of the maxima of the horizontal oar angle. We apply a peak detection algorithm and calculate the stroke time as the distance between the last two maxima. The reciprocal value is the stroke rate.

Stroke Length: We calculate the stroke length as the angle the oar sweeps in the horizontal plane from the catch position to the finish position. We recognize the catch (finish) position by detecting the minimum (maximum) of the horizontal oar angle. We calculate the stroke length from the peak-to-peak amplitude of the horizontal oar angle signal. The actual path length (in meters) that the blade travels in the water depends on the boat/oar settings and is proportional to the angle we calculate.

Ratio Recovery/Drive: The ratio of drive phase to recovery phase is commonly used to characterize the rhythm in rowing. Beginners are particularly advised to keep a ratio of about 2: 1 to assure calm sliding to the catch position and to accelerate the blade through the water to the finish. We detect drive and recovery phase based on the maxima and minima of the horizontal oar angle as indicated in Fig. 4.

Ratio Feathered/Squared: We use a threshold at 45° on the rotation signal to detect the state of the blade for each point in time as indicated in Fig. 4. The point in time to square up the blade is used to characterize the rhythm in rowing. Beginners are particularly advised to square up early to be prepared in time for a good catch.

The algorithms are computationally lightweight to eventually be implemented on wireless sensor nodes with restricted processing resources such as smart phones. The transmission of calculated features rather than raw data reduces the bandwidth requirements.

IV. Application on the Water

We deployed the system in both training and racing conditions. Here we recorded data on the water in skiff racing boats (one person in the boat) from different brands (Stampfli, Filippi, Weidnauer, Empacher). Since rowing technique is individually different we recorded data for 18 participants (age 15-53, weight 59-91 kg, height 174-190 cm). The subjects' rowing experiences ranged from ambitious amateurs to national team rowers (Germany and Switzerland) including current world champions and a current Olympic medalist. For sake of completeness we provide the basic parameters for all 18 participants measured in Table 3. Using the example of two rowers, we demonstrate in detail how our system could integrate into the iterative rowing technique optimization process illustrated in Fig. 1. We have focused on two rowers, Rl (a current Olympic silver medalist) and R2 (a current U23 world champion). Table 3 also includes results for amateur rowers whose rowing technique can be analyzed in the same way. The rowing technique indicators show differences between the amateurs and the more experienced rowers, especially in the training sessions. For both the racing and the training sessions the stroke length of the amateurs is on average shorter (race: 102° < 106°; training: 103° < 109°) and shows a higher standard deviation (race: 2.38° > 2.16°; training: 2.78° > 2.04°).

A. Procedure

For each participant we recorded both a training and a racing session. The training Session comprised 1000 m of rowing with a constant stroke rate between 18 and 21 spin, focusing on good technique. The racing session comprised 1000 m of rowing in racing conditions against another boat aiming to achieve the minimum possible time and with an unlimited stroke rate. Before each session, the rowers were instructed to perform calibration movements used to adapt the system to the boat and mounting settings. Namely, each rower moves to the catch, middle and finish positions and holds each position for seven seconds. Afterwards, the participants were asked whether they felt impaired by the system or could row naturally. The test time for both sessions in total was about 30 minutes per participant. Installing and removing the sensors took an additional 15 minutes per participant. In total we recorded over 10 hours of data.

B. Results and Discussion

All participating rowers stated that they were not impaired by the system and could row naturally. So, we consider the system sufficiently unobtrusive to record realistic data.

Fig. 6 depicts for rower Rl the result of a continuous monitoring of the stroke rate, stroke length, ratio recovery/drive and ratio feathered/squared for both racing and training. Considering the visualization of the racing session, the rower's race strategy can be identified: The race begins with a high stroke rate (40 spm), then the stroke rate drops and remains constant in the steady-state rowing phase (about 30 spm). Finally, the stroke rate increases again for the final sprint. During the high stroke rates the minimum time for the drive phase is limited by the rower's strength and water resistance. To achieve an increased stroke rate the rower usually shortens the recovery time. So, the ratio recovery/drive is lower than for the training session. The relatively reduced time for the recovery phase also causes the ratio feathered/squared to drop. In racing the rower tries to focus on good technique and attempts to execute each stroke in the same way as he did at a reduced stroke rate during training.

Stroke Rate and Stroke Length: Stroke length and stroke rate are two of the main important rowing technique indicators. Fig. 7 depicts the stroke length over the stroke rate for two rowers, Rl and R2, for both training and racing. As in Fig. 6 the stroke length decreases as the stroke rate increases for both Rl and R2. During races a broader range of stroke rates occurs. The race starts and finishes with final sprints and therefore higher stroke rates are achieved during these phases of the race (see Fig. 6). The race phases are represented in clusters and are marked with colours exemp!arily for Rl . Here we observe a linear correlation between stroke rate and stroke length. The corresponding slope is lower for Rl than for R2. This means that Rl can keep a higher stroke length, even at an increased stroke rate. Rl shows a low variance in the stroke length and stroke rate and a high absolute stroke length and maximum stroke rate. In the steady-state rowing phase Rl and R2 follow two different rowing styles: Rl performs longer strokes at a lower stroke rate compared to R2. R2 compensates the shorter stroke length with a higher stroke rate.

Detailed Analysis of Stroke Length: Coaches can use our system to analyze variations in the stroke length in more detail. The variance of the absolute stroke length might result from a variance in the catch position, the finish position or both. Fig. 8 depicts the horizontal oar angle over time for rower R3; this is similar to that in Fig. 4. In this example the variance in the finish position is the main contributor to the variability of the overall stroke length. Based on the output of the system the coach is able to advise the rower to perform dedicated exercises to strengthen the muscles which support a stable finish position. This example demonstrates the benefit of our system for the training process.

Rhythm: Fig. 9 depicts the oar rotation angle over the horizontal oar angle. This visualization supports the coach in analyzing the rhythm and helps to identify rowers who square up early or late. How fast the squaring of the blade takes place is indicated by the distance between the dotted red vertical lines. Also, the stability of the blade in the water during the drive phase and in the air during the recovery phase is characterized. In this example we observe an over-rotation for R2 after extracting the blade from the water.

Boat Acceleration: Fig. 10 depicts the boat acceleration in driving direction over the horizontal oar angle. This allows the coach to analyze the impact of changes in different phases of the rowing cycle on the boat acceleration. In this example, the coach could observe an irregularity on the mid-stroke position of subject R2 during racing conditions. Usually, a rower reaches the maximum acceleration during the middle of the stroke, in this case there is a local minimum. The following Table 3 gives basic rowing technique indicators for all 18 participants for training and the steady-state rowing part during the race. In brackets the standard deviation is given.

Racing Training

Rower stroke stroke recovery/drivefeathered/squared stroke stroke rec very/drivefeathered squared length rate length rate

World-Class Rower

Rl 106.56 28.64 0.90 (0.03) 0.43 (0.01) 114.82 19.75 1.59 (0.00) 0.52 (0.00) (1.54) (1.11) (0.87) (0.46)

R2 101.36 31.56 0.90 (0.03) 0.40 (0.01) 107.50 18.47 1.66 (0.01 ) 0.55 (0.00) (2.29) (0.72) (1 75) (0.42)

R3 107.36 32.11 1.01 (0.05) 0.44 (0.01 ) 108.16 18.37 1.86 (0.01) 0.58 (0.00) (2.99) (1.48) (2.83) (0.62)

R4 104.91 30.51 1.08 (0.05) 0.42 (0.01) 108.34 ^■ 17.36 1.87 (0.01 ) 0.57 (0.00) (2.31) (0.62) (3.19) (0.39)

R5 104.49 32.26 0.96 (0.03) 0.42 (0.01) 109.24 20.78 1.54 (0.01 ) 0.53 (0.00) (3- 14) (1.22) (2.28) (0.68)

R6 107.10 31.90 0.94 (0.03) 0.41 (0.01) 107.26 19.36 1.80 (0.00) 0.55 (0.00) (1.48) (0.63) (1 96) (0.42)

R7 1 12.51 33.15 0.96 (0.06) 0.41 (0.02) 1 13.41 20.41 1.59 (0.00) 0.55 (0.00) (1.65) (2.40) (1.35) (0.39)

R8 107.45 29.11 1.04 (0.04) 0.42 (0.02) 107.58 19.33 1.63 (0.00) 0.52 (0.00) (2.16) (0.80) (2.21) (0.42)

R9 102.09 35.93 0.84 (0.03) 0.33 (0.02) 106.35 19.84 1.58 (0.00) 0.46 (0.00) (1.89) ( 1.51) (1.93) (0.39)

Ambitious Amateurs

RIO 101.23 24.93 1.28 (0.08) 0.44 (0.04) 105.01 19.30 1.58 (0.01) 0.47 (0.00) (3.65) (1.08) (4.25) (0.45)

Rl l 97.13 26.01 1.15 (0.07) 0.48 (0.02) 99.58 17.18 1.71 (0.03) 0.59 (0.00) (2.51) (0.77) (1.36) (0.48)

R12 106.27 31.76 1.00 (0.03) 0.38 (0.01) 104.95 17.59 1.63 (0.01) 0.55 (0.00) (1.41) (1.04) (2.73) (0.60)

R13 103.98 28.56 1.01 (0.04) 0.44 (0.01) 106.06 23.10 1.23 (0.00) 0.47 (0.00) (2.10) (0.49) (2.21) (0.57)

R14 98.43 26.57 1.05 (0.06) 0.44 (0.02) 98.59 24.10 1.15 (0.01) 0.48 (0.00) (2.10) (0.92) (2-21) (0.83)

R15 104.12 25.21 1.07 (0.05) 0.41 (0.02) 98.38 18.09 1.30 (0.01) 0.52 (0.00) (1 -67) (0.79) (2.46) (0.71)

R16 105.67 30.68 0.86 (0.02) 0.38 (0.01) 106.29 20.54 1.33 (0.00) 0.45 (0.00) (2.48) (0.98) (1.36) (0.40)

R17 97.99 26.75 1.07 (0.12) 0.49 (0.03) 108.85 19.52 1.57 (0.03) 0.59 (0.00) (3.95) (1.63) (5.53) (1.02)

R18 107.31 25.65 1.15 (0.07) 0.45 (0.02) 102.84 15.80 1.79 (0.01) 0.61 (0.00) (1.58) (0.91) (2.93) (0.54) V. Conclusion and Outlook

By collaborating with national rowing coaches, we have identified important indicators relating to rowing technique. We have found that complete IMUs or even a subset of these sensors are suitable for measuring a subset of these rowing technique indicators. In a feasibility study we have investigated the optimal sensor setup. Our final sensor setup consists of two IMUs on the oars and one on the boat. For evaluation on the water we have selected 18 participants whose experience level ranged from ambitious amateurs to world- class rowers. For each we have recorded a training and a racing session each of which consisted of 1000m rowing. We have shown that we are able to measure rowing technique indicators such as stroke length and stroke rate for both amateurs and world-class rowers. Using the data derived from two world-class rowers we have demonstrated how our system supports the iterative process of optimizing rowing techniques. In our approach the coach as an expert is in the loop, who integrates the overall context such as the wind, water and boat conditions, body height, or the individual style of the rower concerned to come up with holistic conclusions.

Our next steps are to use wireless DVIUs for easier application and to investigate the influence of potential packet loss and sensor sampling rate. We also plan to monitor the rower's body postures to capture additional rowing technique indicators, to provide condensed information using real-time feedback and to evaluate the benefit of the system in a long-term study.

Our vision is a rowing boat that is equipped with an unobtrusive Rowing Boat Area Network (RBAN) that continuously monitors the rower's technique for both offline analysis and real-time feedback. We see potential in a modular RBAN that seamlessly and comfortably integrates into the rower's equipment. Depending on the requirements a set of dedicated sensors (e.g. on-body or physiological sensors) could monitor rowing performance indicators like physical strength, mental strength, boat stability and synchronicity in crew boats. We believe that taking advantage of the progress in intelligent sensors and communication technology will help to push the boundaries of the sport.

Apart from being applied to rowing sports, the applied techniques to observe rotary angles to improve performance of a user can also be applied to other sports, e.g. by attaching an EVIU to a sports or rehabilitation device, such as a golf club, tennis racket, snowboard, ski or a weapon. It could also be used as input device for computer games or entertainment. References

[1] The perfect stroke www.britishrowing.org, 201 1. [2] FISA Rule Book, 201 1 , http://www.worldrowing.com/fisa/resources/rule-books. [3] W. Fritsch, Das grosse Buch vom Rennrudern. Meyer, 2005. [4] D. Altenburg, K. Mattes, and J. Steinacker, Handbuch Rude rt raining. Limpert, 2008. [5] Virtual Reality Goggles www.biorow.com/PSfiles/VFS3.pdf , 2011. [6] N. Schaffert and K. Mattes, "Sonification of the boat motion to improve the boat run during technique training and racing in rowing," in Cabri, J. et al. Congress of the European College of Sport Science, 2008. [7] V. Kleshnev, "Boat acceleration, temporal structure of the stroke cycle, and effectiveness in rowing," Proceedings of the Institution of Mechanical Engineers, Part P: Journal of Sports Engineering and Technology, vol. 224, no. 1, pp. 63-74, 2010. [8] A. Baca, P. Kornfeind, and M. Heller, "Feedback systems in rowing," The Engineering of Sport 6, pp. 407^412, 2006. [9] R. King, D. Mcllwraith, B. Lo, J. Pansiot, A. McGregor, and G. Yang, "Body sensor networks for monitoring rowing technique," in Wearable and Implantable Body Sensor Networks (BSN), 2009., 2009. [10] R. Anderson, A. Harrison, and G. Lyons, "Accelerometry-based feedback-can it improve movement consistency and performance in rowing?" Sports biomechanics, vol. 4, no. 2, 2005. [11] J. von Zitzewitz, P. Wolf, V. Novakovic, M. Wellner, G. Rauter, A. Brunschweiler, and R. Riener, "Real-time rowing simulator with multimodal feedback," Sports Technology, vol. 1, no. 6, pp. 257-266, 2008.

[12] www.nkhome.com Nielsen Kellermann. [13] A. Sabatini and V. Genovese, "Gyroscopic measurements of the rower's oar pitch angle," in Proceedings of the 7th WSEAS International Conference on Automation & Information. World Scientific and Engineering Academy and Society (WSEAS), 2006, pp. 1-6. [14] F. Gravenhorst, B. Tessendorf, B. Arnrich, and G. Troster, "Analyzing rowing crews in different rowing boats based on angular velocity measurements with gyroscopes," in International Symposium on Computer Science in Sport (IACSS 2011), 2011. [15] J. Wagner, U. Bartmus, and H. De Marees, "Three-axes gyro system quantifying the specific balance of rowing," International journal of sports medicine, vol. 14, pp. 35-35, 1993. [16] P. Sinclair, A. Greene, and R. Smith, "The effects of horizontal and vertical forces on single scull boat orientation while rowing," in ISBS-Conference Proceedings Archive, vol. 1, no. 1 , 2009. [ 17] F. Gravenhorst, B. Tessendorf, and G. Troster, "Towards a rowing technique evaluation based on oar orientation," in International Conference on Pervasive Computing (Pervasive 2011), 201 1. [ 18] D. Bannach, O. Amft, and P. Lukowicz, "Rapid prototyping of activity recognition applications," IEEE Pervasive Computing, 2008. [18] J. Wendel. Integrierte Navigationssysteme: Sensordatenfusion, GPS und Inertiale Navigation. Oldenbourg Wissenschaftsverlag, 2007.

Claims

1. Mobile computing system, preferably with at least part of the system being wearable by a user or attachable to a sport or rehabilitation equipment, comprising a sensor system, characterized in that the sensor system contains at least one gyroscope or equivalent sensor to capture rotary orientation and/or motion of the user or of the Sport or rehabilitation equipment.

2. Mobile computing system according to claim 1, characterized in that the sensor system comprises at least two sensors, wherein the mobile computing system is adapted to calculate a differential signal between at least two of the sensors of the sensor system.

3. Mobile computing system according to claim 1 or 2, characterized in that the sensor system includes as sensors at least one gyroscope and at least one additional inertial measurement sensor, wherein at least one sensor is fixed on a boat and at least one sensor, preferably a gyroscope, is fixed on an oar, and wherein the mobile computing system is adapted to calculate a differential signal between said sensors.

4. Mobile computing system according to any one of claims 1 to 3, characterized in that, as sensor system, one inertial measurement sensor is positioned on a boat and one inertial measurement sensor is positioned on an oar, wherein the sensor on the oar provides orientation data of the oar and the sensor on the boat provides data of the boat's orientation, and wherein the mobile computing system is adapted to calculate a differential signal from signals of the inertial measurement systems to extract oar orientation angles relatively to a boat-fixed coordinate system.

5. Mobile computing system according to any one of claims 1 to 4, characterized in that, as sensor system, two inertial measurement units are positioned on a boat and on an oar, wherein the sensor on the oar provides orientation data and the sensor on the boat provides data of the boat's acceleration, and wherein the mobile computing system is adapted to calculate a differential signal from signals of the inertial measurement systems to extract oar orientation data that is independent of an absolute boat movement.

6. Mobile computing system according to any one of claims 1 to 5, characterized in that at least one sensor is fixed on the sliding seat and at least one sensor is fixed on the boat, wherein the mobile computing system is adapted to calculate a differential signal to extract sliding seat position data relatively to a boat-fixed coordinate system, which means it is independent of an absolute boat movement.

7. Mobile computing system according to any one of the claims 1 to 6, characterized in that the mobile computing system is adapted to convert sensor values of different sensors into one coordinate system in order to compensate for a spatial misalignment between the individual sensors.

8. Mobile computing system according to claim 7, characterized in that the coordinate system is a coordinate system of one of the sensors of the sensor system.

9. Mobile computing system according to any one of claims 1 to 8, comprising further sensors for motion and magnetic field detection, preferably a complete inertial measurement unit (IMU), capturing acceleration, magnetic fields and rotation.

10. Mobile computing system according to any one of claim 1 to 9, further comprising a programmable memory unit, characterized in that the programmable memory unit can be programmed with customized software for a specific sport or rehabilitation activity, so that meaningful information for the specific sport or rehabilitation activity is rendered based on captured sensor data and made available to a recording and/or a user feedback unit, preferably a display unit.

1 1. Mobile computing system according to any one of claims 1 to 10, characterized in that it further comprises a GPS receiver.

12. Mobile computing system according to any one of claims 1 to 1 1 , characterized in that a part of the system, preferably a part comprising a user feedback and/or recording unit, is separated from the sensor system and linked to it via a communications channel, preferably a wireless communications channel.

13. Mobile computing system according to any one of claims 1 to 12, characterized in that at least the part of the system containing the sensor system is attached to or incorporated in a wrist-worn or feet-worn device, preferably a watch.

14. Mobile computing system according to any one of claims 1 to 12, characterized in that at least the part of the system containing the sensor system is attached to or incorporated in a piece of clothes, preferably a shirt, cap or shoe.

15. Mobile computing system according to any one of claims 1 to 12, characterized in that at least the part of the system containing the sensor system is attached to or incorporated in a piece of sports or rehabilitation equipment, preferably an oar or paddle.

16. A method to operate a mobile computing system according to any one of the previous claims, wherein part of the mobile computing system that comprises a sensor system is attached to the user's body or a moveable object or flexible object, preferably to an oar, paddle or to the wrist of a user and wherein the sensor data is interpreted towards meaningful information concerning the performed activity or movement, preferably for rowing or paddling sports.

17. The method according to claim 16, characterized in that a differential signal is calculated between at least two sensors of the sensor system.

18. The method according to claim 17, characterized in that the sensor values are converted into one coordinate system in order to compensate for a spatial misalignment between the individual sensors, wherein, preferably, said coordinate system is a coordinate system of one of the sensors of the sensor system.

19. The method according to any one of claims 16 to 18, characterized in that meaningful information is calculated taking into consideration the rotary orientation information of the oar, paddle or wrist.

20. The method according to any one of claims 16 to 19 and wherein the meaningful information is considered to be taken from the following group of data: i. the trajectory of the horizontal and/or vertical and/or rotation oar/paddle angle j. the timing and angular position of the stroke's start (catch),

k. the timing and angular position of the stroke's end (finish),

I. the stroke length,

m. the stroke rate,

n. the classification of the stroke phase (drive or recovery),

0. the ratio drive versus recovery,

p. the timing and angular position of the blade's squaring or feathering movement,

1. the timing and angular position of the blade when it dips into the water or leaves the water.

21. The method according to any one of the claims 16 to 20, characterized in that the method involves a step that evaluates the meaningful information against a quality of the user's movement and/or suggests changes in his movement.

22. A method to operate a plurality of mobile computing systems, each according to any of the claims 12 through 15, wherein each mobile computing system comprises a storage and/or communication unit and data from all mobile computing systems is collected and arranged in a synchronized way, which means that the sensor data from one individual mobile computing system of the plurality of mobile computing systems can be compared with the sensor data which has been acquired at the same time from another mobile computing system of that plurality of mobile computing systems.

23. The method to operate a plurality of mobile computing systems according to claim 22, wherein each of the mobile computing systems comprises additionally a GPS receiver, characterized in that the captured data from the sensor system is linked to a received GPS system time before being stored and/or transmitted.

The method to operate a plurality of mobile computing systems according to claim 22 or 23, wherein each of the mobile computing systems is attached to the wrist, oar or paddle of a team member of a rowing or paddling team, the individual mobile computing systems are operated according to a method according to any of the claims 16 through 21 and the collected meaningful information is evaluated towards synchrony between team members.