Exercise repetition detection for resistance training based on smartphones

Exercise repetition detection for resistance training based on smartphones,

+ Pers. Ubiquit. Comput.(2013),

- Igor Pernek et al,

/ 맹욱재

x 2015 Rainy Fall

UX Lab, Graduate School of Convergence Science and Technology, Seoul National Univ. 2

Contents

1. Intro

2. Related Works

3. System & Algorithm

4. Evaluation


Intro - Why this paper?

WookFit : Wearable Device for Breast Cancer Survivors doing Resistance ExcersiceElastic-band stretching Counter


Intro - Why this paper?(Cont.)

Key word for searching paper

Breast Cancer Rehabilitations

Resistance Excersise(traning)

Home-based Excersise(traning)

Lymphedema

Theraband(Resistance band)

Wearable

IoT

Quality of Life

Smartphone Application

Accelerometer


Key word for searching paper

Breast Cancer Rehabilitations

Resistance Excersise(traning)

Home-based Excersise(traning)

Lymphedema

Theraband(Resistance band)

Wearable

IoT

Quality of Life

Smartphone Application

Accelerometer

Intro - Why this paper?(Cont.)


Intro - Take away

1. Dynamic Time Warping algorithm can accurately count the number of repetition, duration.

2. Giving real time feedback on Quality of Performance can be done using duration, start, end time of repetition


Intro - Abstract

Problem

Solution

Most of systems focus only on cardio exercises such as running and cycling

Exploit smartphones to support leisure activities with a focus on resistance training.

Off-the-shelf smartphones without additional external sensors can be leveraged to capture resistance training data and to give reliable training feedback

Algorithm

Dynamic Time Warping based algorithm to detect individual resistance training repetitions from the smartphone’s acceleration stream


Intro - Abstract

Feedback in detail

Providing feedback about the quality of repetitions, use the duration of an individual repetition and analyze how accurately start and end times of repetitions can be detected by our algorithm

In terms of the number of correctly recognized repetitions. 3,598 repetitions performed by 10 volunteers exercising in 2 distinct scenarios a gym and a natural environment.

The results show an overall repetition miscount rate of about 1 % overall temporal detection error of about 11 % of individual repetition duration.

Evalution


Related Works

Manual input problem

manual input is cumbersome, taking notes about resistance training is often avoided, negatively affect people’s activity levels and decrease their motivation for regular training

https://www.jefit.com/products/androidpro/index.php https://play.google.com/store/apps/details?id=com.namuan.health&hl=ko

https://www.jefit.com/products/androidpro/index.php

https://play.google.com/store/apps/details?id=com.namuan.health&hl=ko


Related Works - Fitlinxx

https://www.youtube.com/watch?time_continue=303&v=BfnPogUokKs

1. very expensive, as the gym has to be equipped with specific exercise machines and infrastructure, and has therefore not gained broader popularity 2. have to log in to the system prior to performing any exercise,

https://www.youtube.com/watch?time_continue=303&v=BfnPogUokKs


Related Works - Recognizing Upper Body Postures using Textile Strain Sensors


Related Works - The Virtual Trainer: Supervising Movements Through a Wearable Wireless Sensor Network

the algorithm relies on exercise specific features, such as the position of the elbow during the biceps curl exercise, and can therefore not be easily applied to a general resistance training exercise.


Related Works - Implementation and evaluation of the personal wellness coach

The acceleration stream-laptop computer threshold-based low-pass filtering algorithm: each transition from resting to exercise state to count the number of repetitions. the number of time units in each state is analyzed to detect struggling during exercise execution. all the repetitions have a similar single peak acceleration footprint, which is not true for exercises performed in different environments


Related Works - Tracking Free-Weight Exercises

feature phone + 2 external accelerometer sensors. data from a glove & chest belt, only derive info for a limited set of free-weight exercises. simple peak detection algorithms that are no detect a repetition’s start and end time ->lack capabilities for providing qualitative feedback based on the duration of repetitions.


Related Works - myHealthAssistant: A Phone-based Body Sensor Network that Captures the Wearer’s Exercises throughout the Day

advance ‘Tracking Free-Weight Exercises' by making the repetition detection algorithm robust to different exercising speeds


Related Works - What can an arm holster worn smart phone do for activity recognition?

arm worn smartphones to recognize a number of upper body resistance training exercises from a continuous acceleration stream. Main focus lies on spotting series of exercises, repetitions- peak detection. start & end times of individual repetitions cannot be detected.


Related Works

Missing information

1) not ubiquitous -> mobile(smartphone)

2) targeted to a set of specific exercises -> broader set of exercises, performed using different types of equipment

3) not support fine-grained tracking of individual repetitions. important to give feedback about the correctness and quality of exercising-> by detection of duration of individual repetitions.

WookFit

1) mobile(smartphone) + Wearable device 2) targeted to a set of specific exercises 3) not support fine-grained tracking of individual repetitions.


System - Requirement

1) Should not add high extra cost nor demand the cumbersome installation of additional sensors and tools => off-the-shelf smartphones

2) Easy to use interface, be encouraging by suggesting particular exercises, and give feedback on the exercising performance as well as the training progress=> Quality of Performance Feedback

3) Usable in different environments including gyms and natural environments and cover a wide range of exercises=> Constrained(gym, fitness center) + Unconstrained(free-weight, resistance band, body weight)


System - Resistance Training Assistant


System - Resistance Training Assistant

1) Download different training plans and exercises from an online training database.

2) The application proposes exercises to perform along with some additional information - the number of repetitions required or the intensity(weight)

3) Realtime feedback about the desired exercise effect by evaluating the duration of the repetitions.

4) Permanently stores the number and the intensity of repetitions for each exercise and allows users to upload it to an online profile to track progress and observe exercising trends (information may be shared within an online community).


Unconstrained Exercise Constrained Exercise

System - Device Position


System - Selected Exercise


Algorithm - Overview


Algorithm - Dynamic Time Warping

Definition

In time series analysis, algorithm for measuring similarity between two temporal sequences which may vary in time or speed

https://en.wikipedia.org/wiki/Time_series_analysis

https://en.wikipedia.org/wiki/Algorithm


Evaluation - Precision, Recall, F score


Evaluation - Box plot for Temporal error of s,e

Fig. 5 Temporal error of detecting individual repetition start and end times normalized by the total duration of the repetition for externally generated repetition patterns (median, quartiles, and extreme values)

감사합니다

맹욱재 [email protected]

mailto:[email protected]?subject=

Exercise repetition detection for resistance training based on smartphones

Devices & Hardware

Transcript of Exercise repetition detection for resistance training based on smartphones