Wearable Technology and Visual Reality Application for Healthcare Systems
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
electronics Article Wearable Technology and Visual Reality Application for Healthcare Systems Kuang-Hao Lin * and Bo-Xun Peng Department of Electrical Engineering, National Formosa University, Yunlin 632, Taiwan; 11065131@gm.nfu.edu.tw * Correspondence: khlin@nfu.edu.tw Abstract: This study developed a virtual reality interactive game with smart wireless wearable technology for healthcare of elderly users. The proposed wearable system uses its intelligent and wireless features to collect electromyography signals and upload them to a cloud database for further analysis. The electromyography signals are then analyzed for the users’ muscle fatigue, health, strength, and other physiological conditions. The average slope maximum So and Chan (ASM S & C) algorithm is integrated in the proposed system to effectively detect the quantity of electromyography peaks, and the accuracy is as high as 95%. The proposed system can promote the health conditions of elderly users, and motivate them to acquire new knowledge of science and technology. Keywords: wearable; IoT; VR; EMG; healthcare 1. Introduction With the advances of modern medicine, people’s average lifespan has prolonged; hence, the aging of population structures has become a global issue. Moreover, the changes Citation: Lin, K.-H.; Peng, B.-X. of lifestyles have induced various chronic diseases, increasing the demand for medical Wearable Technology and Visual care. As routine health information can enhance disease prevention and medical care, Reality Application for Healthcare the application of telecare in routine health care can promote people’s quality of life. Systems. Electronics 2022, 11, 178. In telecare, the first-aid devices can also collect personal physiological information for https://doi.org/10.3390/ ambulances or relevant medical units, so that the patients under care can receive instant electronics11020178 and effective assistance. Literatures [1,2] used electrocardiography (ECG) combined with the Internet of Things Academic Editor: Stefanos Kollias (IoT) to design a remote monitoring and telemetry system that can improve the accuracy Received: 29 November 2021 and reliability of heart health monitoring. Literatures [3,4] used electroencephalography Accepted: 5 January 2022 (EEG) to design a sleep monitoring system and an epilepsy prediction system. Bhowmick Published: 7 January 2022 used photoplethysmography (PPG) to design a PPG monitoring system and stored the PPG Publisher’s Note: MDPI stays neutral model parameters in the cloud for clinical diagnosis [5]. Huang used electromyography with regard to jurisdictional claims in (EMG) to design a diet monitoring system, which receives EMGs of the mastication muscles published maps and institutional affil- through glasses to detect intake-related events [6]. iations. This study combined physiological signal modules (namely, ECG, EEG, PPG, and EMG) with innovative educational technologies, such as the Internet, IoT [7,8], and virtual reality (VR) [9,10], to develop a cloud system with telecare services [11,12]. Using the VR and open-source software (OSS), this study developed an interesting VR digital game that Copyright: © 2022 by the authors. could stimulate the elderly users to utilize the application for health management [13–15]. Licensee MDPI, Basel, Switzerland. The usage scenario of the proposed VR health game is shown in Figure 1. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/). Electronics 2022, 11, 178. https://doi.org/10.3390/electronics11020178 https://www.mdpi.com/journal/electronics
Electronics 2022,11, Electronics2022, 11,178 x FOR PEER REVIEW 22 of of 17 17 Figure 1. Schematic diagram of a system usage scenario. Figure 1. Schematic diagram of a system usage scenario. Artificial intelligence (AI) has been widely used in the identification and analysis of Artificialsignals physiological intelligence in recent(AI)years. has beenRecentwidely used in common EMGthe analyses identification and analysis and studies focused of physiological on AI application. signals Forin recent years. example, SugiartoRecent [15] common analyzed EMG analyses the surface EMG and studiesand (sEMG) focused high- on AI application. density EMG (HD-EMG) For example, Sugiarto of the neck muscle [15](designated analyzed the surfaceand muscle) EMG used(sEMG) and high- Convolutional density EMG (HD-EMG) of the neck muscle (designated muscle) Neural Network (CNN) for training to correct the end-to-end delay of the VR system and and used Convolutional Neural Network alleviate negative (CNN) effects suchfor training as motion to correct sickness. the end-to-end Sugiarto delay [15] also of the VR employed system three pairsand of alleviatesEMG wireless negative effects sensors from such as motion Delsys Trignosickness. Sugiarto (Delsys, MA, USA)[15] alsosoftware. as the employed three pairs Pancholi [16] of wireless analyzed thesEMG peak sensors averagefrom power Delsys (PAP)Trigno of EMG (Delsys, MA, USA) and adopted Linearas Discriminant the software. Analysis Pancholi [16] analyzed (LDA) the peakDiscriminant and Quadratic average power (PAP) of Analysis EMGfor (QDA) andtraining adoptedand Linear Discriminant prosthetics. The Analysisof analysis (LDA) and [16] Pancholi Quadratic has been Discriminant done usingAnalysis MATLAB (QDA) 2015afor ontraining an i7 core.andRaurale prosthetics. [17] The analysis of Pancholi [16] has been done using MATLAB 2015a employed subsequently classified to classify and identify eight kinds of EMG actions. The on an i7 core. Raurale [17] employed system subsequently is also shown to operate classified to classify in real-time on anandARM identify Cortexeight A-53kinds of EMG embedded actions. processor The system suitable is also shown for housing in anto EMGoperate in real-time wearable device.onAdditionally, an ARM Cortex Wang A-53 embedded [18] modifiedpro- the So andsuitable cessor Chan algorithm for housing to inincrease an EMG thewearable accuracydevice. of ECG peak detection Additionally, Wang to [18] 99.16% from modified 94.61% andChan the So and used algorithm FPGA for to verification. increase the However, accuracyitofisECG complicated and nottoapplicable peak detection 99.16% from to biomedical 94.61% andVR used games, FPGAwhich are characterized for verification. However, by low it is complexity complicated[19,20]. and not Therefore, applicable thisto study aims to develop an innovative realization architecture biomedical VR games, which are characterized by low complexity [19,20]. Therefore, this with low complexity. The proposed study aims system specifically to develop targets therealization an innovative irregularities of EMG peaks, architecture with low in order to improve complexity. The the operational proposed system accuracy and smoothness specifically of VR games.ofToEMG targets the irregularities verify the EMG peaks, peaktodetection in order improve accuracy, this study the operational first used accuracy andthe automaticof smoothness calibration VR games. detection To verify (ACD) the EMG to capture the EMC peak detection signals, accuracy, and setstudy this a threshold parameter first used to assess the the automatic captureddetection calibration signals. To further (ACD) toimprove capture thethe EMG peak detection accuracy, this study applied the average slope EMC signals, and set a threshold parameter to assess the captured signals. To further im- maximum So and Chan (ASM prove Sthe& C)EMGalgorithm [21–25]. accuracy, peak detection The comparisonthis studyof recent applied studies on physiological the average slope maximumsignal detection systems is as shown in Table 1. The operation sensitivity So and Chan (ASM S & C) algorithm [21–25]. The comparison of recent studies on physi- was thus enhanced in the proposed VR game. ological signal detection systems is as shown in Table 1. The operation sensitivity was thus enhanced Table 1. Comparison in the proposed of Recent VR on Studies game. Physiological Signal Detection Systems. Table 1.Refrence Comparison of Recent Studies on Physiological Signal Algorithm Detection Systems. Complexity Equipment Refrence Algorithm Complexity Delsys Trigno Equipment Sugiarto [15] CNN High IMU Delsys Trigno Sugiarto Pancholi[15] [16] CNN LDA High High I7 core IMU Raurale[16] Pancholi [17] Subsequently LDA classified Medium High ARMI7Cortex core A-53 Wang [17] Raurale [18] Enhanced classified Subsequently So and Chan Low Medium FPGA A-53 ARM Cortex Wang [18] Our propose Enhanced So ASM and S&CChan LowLow FPGA nRF52840 Our propose ASM S&C Low nRF52840
Electronics 2022, 11, x FOR PEER REVIEW 3 of 17 Electronics 2022, 11, 178 3 of 17 The remainder of the paper is organized as follows. Section 2 describes the system architecture design and The remainder its specifications. of the Section paper is organized as 3follows. discusses the detailed Section design 2 describes themethod- system ology for EMG peak detection. Section 4 presents the analysis of the experimental results. architecture design and its specifications. Section 3 discusses the detailed design methodol- Finally, ogy for conclusions are drawnSection EMG peak detection. in Section 5. 4 presents the analysis of the experimental results. Finally, conclusions are drawn in Section 5. 2. System Architecture 2. SystemArchitecture Hardware Architecture Hardware Architecture Figure 2 shows the hardware architecture of the smart wearable biomedical sensing systemFigure 2 shows developed inthe thishardware study. The architecture sensing end of the usessmart wearableanbiomedical the Tri-BLE, sensing IoT development system developed in this study. The sensing end uses the Tri-BLE, platform, as the main body, and is provided with a Tri-EMG sensor, a button, and an an IoT development platform,light. indicator as theThe main body, human andsignal EMG is provided is sent with via a adisposable Tri-EMG electrode sensor, a to button, and an the Tri-EMG indicator light. The human EMG signal is sent via a disposable electrode sensor, and subsequently exported. The Tri-BLE platform filters the received EMG data, to the Tri-EMG sensor, and the and subsequently threshold of the EMGexported. signal The Tri-BLE is judged platformtofilters according the received the relationship EMG data, between the and the threshold of the EMG signal is judged according to the relationship EMG signal and the threshold. The output state is transferred to the central controller and between the EMG signal and the threshold. The output state is transferred to the central processed. Furthermore, the button controls a calibration mechanism. The threshold of controller and processed. the EMG signalFurthermore, varies withthethe button human controls body aandcalibration mechanism. the electrode The position, andthreshold of the this threshold EMG signal varies with the human body and the electrode position, and this can be adjusted according to the indicated action after the button is pressed. The indicator threshold can be adjusted according to the indicated action after the button is pressed. The indicator light light displays the state of the sensing end—the light is red during calibration, and the light displays the state of the sensing end—the light is red during calibration, and the light turns turns green if the EMG signal exceeds the threshold. The central controller integrates the green if the EMG signal exceeds the threshold. The central controller integrates the states of states of the left- and right-hand sensing ends, and communicates with the VR device to the left- and right-hand sensing ends, and communicates with the VR device to control the control the action in the game. Additionally, the system has a mobile app that is linked action in the game. Additionally, the system has a mobile app that is linked via Bluetooth. via Bluetooth. The muscle conditions of the left and right hands can be recorded, and ac- The muscle conditions of the left and right hands can be recorded, and accessed by doctors cessed by doctors for future diagnoses. for future diagnoses. Figure Figure 2. 2. Hardware Hardware architecture architecture of of the the smart smart wearable biomedical sensing wearable biomedical sensing system. system. Table 2 lists the equipment specifications. specifications. Tri-BLE Tri-BLE and and Tri-EMG Tri-EMG ofof TriAnswer TriAnswer series are used for the biomedical development development platform, platform, Arduino Arduino Uno Uno is is used used as as the central includes an HTC VIVE VR headset controller, and the monitor set includes headset and and an app, with the associated equipment of HTC-VIVE and Android smartphone, respectively. respectively. Table 2. Equipment specifications. Table 2. Equipment specifications. System System Requirement Requirement Equipment Equipment Biomedical Biomedical TriAnswer TriAnswer Development Development Platform Platform (Tri-BLE, (Tri-BLE, Tri-EMG) Tri-EMG) Central controller Central controller Arduino ArduinoUNO UNO Virtual reality HTC-VIVE Cosmos Monitor Virtual reality HTC-VIVE Cosmos Monitor Android Smartphones Android Smartphones
Electronics 2022, 11, x FOR PEER REVIEW 4 of 17 Electronics 2022, 11, 178 4 of 17 3. EMG Peak Detection 3. EMG 3.1. System Peak FlowDetection 3.1. Figure System 3Flow presents the moving average system flowchart. Notably, the human EMG signalsFigure 3 presents the are complicated andmoving average susceptible systeminfluences. to external flowchart.This Notably, studythe human thus EMG employed signals are complicated and susceptible to external influences. This the moving average and ACD methods to reduce the number of signal judgment errors.study thus employed thethe For moving moving average andthe average, ACD EMG methods signals to arereduce importedthe at number 500 Hz.ofThe signal judgment length of eacherrors. data For the moving average, the EMG signals are imported at 500 Hz. The length is 8 bits, and the system output is the average of 20 data points. Therefore, the first of each data in first is 8(FIFO) out bits, and the system buffer which output is the can store average 20 data of 20was points data points. used. Therefore, When the first the buffer was in first filled out (FIFO) with 20 databuffer which points, can storevalue the average 20 data waspoints was used. calculated as theWhen outputtheY(n). buffer was filled with 20 data points, the average value was calculated as the output Y(n). X(n) FIFO Buffer No n>20 Yes Average FIFO Buffer Y(n) Figure Figure3.3.Moving Movingaverage averagesystem systemflowchart. flowchart. Figure Figure44isisthe the ACD ACD system flowchart. flowchart. The The ACD ACDsystem systemadopts adopts10,000 10,000EMG EMGsignals signalsas asthe thethreshold thresholdcalibration calibrationcriteria. criteria.First, First, the the input data are compared compared withwiththe theeight eightdata data ininthe thebuffer. buffer.When Whenthe theinput inputis is larger larger than than thethe buffer, buffer, thethe minimum minimum value value in the in the buffer buffer is is replaced replaced by bythethe new new input input data. data. After After thethe system system comparesthe compares the10,000 10,000data datapoints, points,the the threelargest three largestdata data points points are are removed from from the the buffer buffertotoavoid avoidthethesignal signalerrors errorsresulting resultingin ina high a high threshold. threshold. After Afterfivefive data points data are are points averaged, this this averaged, value is subtracted value by thebyinitial is subtracted the value (O initial value ) and multiplied by the scale factor of D to obtain the threshold after ini (Oini) and multiplied by the scale factor of D to obtain the threshold after cal- calibration. Figure 5 is the average slope maximum So and Chan (ASM S & C) system flowchart. ibration. The initial_value is calculated first, and if the EMG signal exceeds the initial_value, the ini- tial_slope_maxi is being searched. The slope_threshold is then obtained by the initial_slope_maxi. If two consecutive EMG slopes exceeds the slope_threshold, the current signal is recorded as the starting point of the waveform characteristics, and the peak could be determined. Finally, the peak and starting points are used for updating the maxi. The ASM S & C algorithm is described in Section 3.2. 3.2. Algorithm EMG signal shows the potential difference when a muscle contracts. According to the state of motion, the amplitude and frequency of the electromyogram change accordingly. Under ideal conditions, an electromyogram indicates no movement, while obvious fluc- tuations occur when the muscles are contracted (as shown in Figure 6). However, EMG signals are susceptible to random interference caused by radio waves, electrode noise, and other factors [26–28].
Y(n) No Yes n>10,000 Electronics 2022, 11, x FOR PEER REVIEW 5 of 17 Electronics 2022, 11, 178 5 of 17 Compare Average Y(n) Scale Factor Buffer[0:7] No Yes D n>10,000 Compare Initial AverageValue Oini Scale Factor Buffer[0:7] threshold D Figure 4. ACD system flowchart. Initial Value Figure 5 is the average slope maximum So and Chan (ASM S & C) system flowchart. Oini The initial_value is calculated first, and if the EMG signal exceeds the initial_value, the ini- tial_slope_maxi is being searched. The slope_threshold is then obtained by the ini- tial_slope_maxi. If two consecutive EMG slopes exceeds the slope_threshold, the current sig- nal is recorded as the starting point of thethreshold waveform characteristics, and the peak could be determined. Finally, the peak and starting points are used for updating the maxi. The ASM S & C algorithm is described in Section 3.2. Figure 4. ACD system flowchart. Figure 4. ACD system flowchart. Figure 5 is the average Y(n) slope maximum So and Chan (ASM S & C) system flowchart. The initial_value is calculated first, and if the EMG signal exceeds the initial_value, the ini- tial_slope_maxi is being searched. The slope_threshold is then obtained by the ini- tial_slope_maxi.No If two consecutive maxi Yes slopes exceeds the slope_threshold, the current sig- EMG nal is recorded as the starting point of the waveform characteristics, and the peak could be determined. Finally, the peak and starting points are used for updating the maxi. The slope_threshold ASM SAverage & C algorithm is described in Section 3.2. (ST) Y(n) initial_value slope No Yes maxi Get No slope > ST initial_slope_maxi slope_threshold Average (ST) Yes update Get maxi EMG Peak initial_value slope Figure5.5.ASM Figure ASMSS&&CCsystem systemflowchart. flowchart. Get No slope > ST initial_slope_maxi Yes update Get maxi EMG Peak Figure 5. ASM S & C system flowchart.
3.2. Algorithm EMG signal shows the potential difference when a muscle contracts. According to the state of motion, the amplitude and frequency of the electromyogram change accord- ingly. Under ideal conditions, an electromyogram indicates no movement, while obvious fluctuations occur when the muscles are contracted (as shown in Figure 6). However, Electronics 2022, 11, 178 6 of 17 EMG signals are susceptible to random interference caused by radio waves, electrode noise, and other factors [26–28]. (A) (B) Figure 6. EMG Figurecomparison with and without 6. EMG comparison with andmovement. (A) No action. without movement. (A) No(B)action. Action.(B) Action. The two Theconsiderations of noise interference two considerations and hardware of noise interference design allow and hardware designtheallow moving the moving average algorithm to effectively average algorithm eliminateeliminate to effectively random interference in a largein random interference amount a largeof data. of data. amount The moving Theaverage moving algorithm used in this average algorithm usedpaper adopted in this paper feedback control ascontrol adopted feedback the filtering as the filtering method. method. After After a fixed a fixed period sizeperiod sizethe is given, is given, outputthe output is the averageis theof average the inputofvalues the input in values the periodin size. the period size. Its Its operating operating principle principle is to remove is tooldest the remove the oldest element element and add the nextand add the nextbefore input value input value beforethe averaging averaging output. the Thisoutput. outputThis valueoutput value divides divides the random the noise random noise evenly andevenly makesandit makes smaller, it thus smaller, thus the making making overallthedata overall data waveform waveform smoother. smoother. The moving The average moving algorithm average algorithm is expressed is expressed in the following in the following equation:equation: ∑ ( P−= ) h i = ∑ P = 0T − 1 X (n − P) (1) Y = (1) T where Y is the output of the moving average; T is the period length; and X(n) is the current where Y is the output of the moving average; T is the period length; and X(n) is the current input data. input data. Although the moving average can effectively remove random noise, EMG changes Although can be easily read the moving and compared withaverage can effectively the threshold remove random signal. However, factors noise, such as EMGdif- changes ferent users or patch positions can change the EMG amplitude and cause a large gap in such as can be easily read and compared with the threshold signal. However, factors different the threshold. users orthis Therefore, patch positions paper designedcan an change the EMG automatic amplitude calibration and cause detection (ACD) a large gap algorithm to solve the problem of threshold error. The ACD algorithm collects 10,000 (ACD) in the threshold. Therefore, this paper designed an automatic calibration detection pieces ofalgorithm data to settothesolve the problem threshold. The ACDof threshold algorithm error. The ACD equation is asalgorithm follows: collects 10,000 pieces of data to set the threshold. The ACD algorithm equation is as follows: 1 ℎ ℎ = ( 1 )n =× k + (1 − ) × (2) " # threshold = kn=1∑ S(n + 3) × D + [(1 − D ) × Oini ] (2) where S is the output after comparison, k = m − 3, D is the scale factor, and Oini is the initial EMG value. where A method S is thesimilar outputtoafterthe sorting comparison,algorithmk=m is − used3, Dinisthethecomparison. scale factor,Asand long Oini is the as the input value of the ACD algorithm is greater than the value in initial EMG value. A method similar to the sorting algorithm is used in the comparison. the buffer, it can be directly replaced. As long asThis the method input valuecan find of thethe ACD largest m data is algorithm and sort them greater than from the largest the value in the buffer, it to the smallest. In orderreplaced. can be directly to prevent Thisnoise methodinfluence, can find thethelargest largest three m datapieces andofsort data themare from the removedlargest from the m data. to the There smallest. Inare ordersignal bounces to prevent occasionally noise influence,inthe thelargest process of EMG three pieces of data are removed signal sensing; hence,fromthe top m data. thethree There of the eightarepieces signal of bounces data are occasionally removed toinprevent the process the of EMG values ofsignal sensing; high signal hence,from bounces the top three of affecting thethe eight pieces average value.of data are removed to prevent the Thevalues of high signal data averaged bouncesbyfrom is multiplied affecting the scale factor theDaverage to improve value. the accuracy of dis- crimination. Finally, Oini is added as the initial compensation of theDthreshold. The data averaged is multiplied by the scale factor to improve The thescale accuracy of factor D discrimination. can obtain the best Finally, Oini is added parameters after as thethe initial compensation numerical of the threshold. statistical analysis, as shownThe scale in Figurefactor D can 7. When theobtain scale the bestisparameters factor 50%, the accuracyafter theofnumerical the EMGstatistical analysis, peak detection canas shown reach 71%.in Figure 7. When the scale factor is 50%, the accuracy of the EMG peak detection can reach 71%.
Electronics 2022, 11, x FOR PEER REVIEW 7 of 17 Electronics 2022, 11, 178 7 of 17 80 60 Accuracy (%) 40 20 0 10 20 30 40 50 60 70 80 90 100 Scale factor D (%) Figure Figure 7. 7. Scale Scale factor factor statistical statistical analysis. analysis. The optimal scale factor in the ACD algorithm varies with the users; therefore, the data from each individual user should be statistically analyzed, and the scale factor cannot be corrected during real-time hardware hardware variations. variations. As a result, the accuracy varies by each user. In In order orderto toenhance enhancethetheaccuracy accuracyinin detection, detection,thisthis study detected study the the detected EMG peaks, EMG and peaks, usedused and the transient peakpeak the transient when the muscle when put put the muscle forthforth strength as the strength as reference frame the reference for the frame for overall the threshold. overall ThisThis threshold. study adopted study thethe adopted So So andandChan Chan (S & (S C) algorithm, & C) which algorithm, which is is the R the wave detection algorithm proposed by So and Chan in 1997 [22], and modified R wave detection algorithm proposed by So and Chan in 1997 [22], and modified it into it into the average the slope average maximum slope maximumSo and Chan So and (ASM Chan S & SC)&algorithm. (ASM C) algorithm. The S & C algorithm is given below The S & C algorithm is given below [22]: [22]: Y(n) is Y(n) is the the processed processed EMG EMG amplitude, amplitude, and the slope and the slope ofof each each point point can can be be found found inin the the time domain time domain by by Equation Equation (3): (3): ( ) slope(n= )=−2 ( −2Y (−n 2) − 2−) ( − Y (−n 1) − 1+) ( + Y (+n1) ++ 1) 2 ( + 2Y (+n2) + 2) (3) The slope_threshold can be The slope_threshold obtained can by Equation be obtained (4): (4): by Equation ℎ ℎ _ _ ℎ ℎ = threshold_paprameter× (4) slope_threshold = 16 × maxi (4) 16 First, a sample quantity is set up according to the sample rate of signal reception to First, find the a sample quantity initial_slope_maxi. Theissample set uprate according used intothis thepaper sample rateHz, is 500 of signal reception as shown in Al- to find the initial_slope_maxi. gorithm 1: S & C algorithm. The sample rate used in this paper is 500 Hz, as shown in Algorithm 1: S & C algorithm. Algorithm Algorithm 1:1:SS&&C C Algorithm. Algorithm. Input: Input: Y(n) Y(n) Output: Output: 1:1: slope slope = 0= 0 3:2: initial_slope_maxi initial_slope_maxi =0= 0 4:3: for for i: 1= to i: = 1 to500500 do do 5:4: if i if≥i3>=then 3 then 6:5: = −2 =×−2 slopeslope Y(i×−Y(i 2) −−2) Y(i− − Y(i1)−+1) Y(i+ +Y(i 1) + 2 ×+ Y(i + 1) 2 ×+Y(i 2) + 2) 7:6: if slope > initial_slope_maxi if slope then then > initial_slope_maxi 7: initial_slope_maxi = slope 8: initial_slope_maxi = slope 8: end if 9: end if 9: end if 10: 10: end forend if 11: end for 11: maxi = initial_slope_maxi 12: maxi = initial_slope_maxi The slope_threshold is determined by using the initial maximum slope. When two consecutive slopes are larger The slope_threshold than theby is determined slope_threshold, using the initial thismaximum point is marked as thetwo slope. When starting con- secutive slopes are larger than the slope_threshold, this point is marked as the starting point
Electronics 2022, 11, x FOR PEER REVIEW 8 of 17 Electronics 2022, 11, 178 8 of 17 of the signature waveform, and then the peak point of waveform could be determined. The pointmaxi is signature of the then updated, as expressed waveform, inthe and then Equation (5): of waveform could be determined. peak point _ The maxi is then updated, as expressed − (5): in Equation = + (5) _ f irst_maxi − maxi maxi = + maxi (5) f iliter_parameter _ = Height of p_point − Height of start_point (6) f irst_maxi = Height of p_point − Height of start _point (6) For the filiter_parameter which can be set as 2, 4, 8, or 16, this paper selected 16. The peak For wasthe recorded, and the which filiter_parameter peak detection can be setaccuracy as 2, 4, 8, of or the algorithm 16, this was calculated paper selected [22]. 16. The peak was recorded, There and the were 100,000 EMGpeak detection samples, theaccuracy number of the algorithm making was 154, a fist was calculated SP was[22]. the There num- were ber of100,000 EMG samples, correct peaks captured the number ofMP successfully, making was the a fist was 154, number SP waspeaks of correct the number that wereof correct peaks missed, and EP captured was thesuccessfully, MP waserror. position of capture the number of correct The accuracy canpeaks that were be computed bymissed, Equa- and (7). tion EP was The the S &position of capture C algorithm error. captured 154The accuracy peaks can be computed successfully, by Equation but there were (7). 54 capture The S & C algorithm captured errors, for an accuracy of 74%. 154 peaks successfully, but there were 54 capture errors, for an accuracy of 74%. SP Accuracy(%) = × 100 (7) SP + MP SP+ EP Accuracy(%) = × 100 (7) Figure 8 simulates the peak detection of + MP SPthe S&+ EP C algorithm. It was observed that the S & CFigure algorithm had no the 8 simulates obvious peak waveform detection of characteristics in the first the S & C algorithm. 500 samples, It was observedindi- that cating the S &there were errors C algorithm hadinnocapturing the initial_slope_maxi obvious waveform and in characteristics leading to errors the first in the 500 samples, indicating there subsequent were errors slope_threshold in capturing and the initial_slope_maxi and leading to errors in the peak detection. subsequent slope_threshold and peak detection. 2 EMG value EMG Peak 1.9 1.8 Voltage (Vp-p) 1.7 1.6 1.5 1.4 1.3 0 10 20 30 40 50 60 Time(s) Figure Figure 8. 8. Peak Peak detection detection of of the the SS & &CC algorithm. algorithm. The S & C C algorithm algorithm is is mainly mainly used used for for ECG ECG detection. detection. TheThe main main difference difference between between EMG and ECG signalssignals is is that that an an EMG EMG signal signal has has obvious obvious waveform waveform characteristics characteristics only if the muscle muscle contracts, contracts, whereas whereas an an ECG ECG signal signal always always hashas waveform waveform characteristics. characteristics. Hence, Hence, the initial_slope_maxi capture requirements of the the initial_slope_maxi capture requirements of the S & algorithm S & C algorithm are not applicable to applicable EMG. The EMG. initial_slope_maxi in The initial_slope_maxi in the the SS & &C C algorithm algorithm is is captured captured in in aa preset period of cycle In ECG time. In ECG with with aacycle cyclethat thatchanges changesslightly, slightly, accurate initial_slope_maxi can be accurate initial_slope_maxi be obtained obtained from the S & C algorithm after adjusting from adjusting thethe cycle cycle based based on the ECG signals signals of different persons. However, However,for forEMG EMGsignals signalswith irregular with irregular changes, changes,detection detectionbased on a on based fixed cycle a fixed is not feasible. In order to solve this problem, the capture requirements cycle is not feasible. In order to solve this problem, the capture requirements of the maxi- of the maximum slope slope mum were changed in thisinpaper, were changed as shown this paper, in Algorithm as shown 2: ASM in Algorithm 2: SASM & C Salgorithm. & C algorithm.
Electronics 2022, 11, 178 9 of 17 Algorithm 2: ASM S & C Algorithm. Input: Y(n) Output: 1: initial_value = avg(sum(Y(1:n))) × (λ + 1) 2: slope = 0 3: initial_slope_maxi = 0 4: count = 0 5: en = 0 6: for i:= 1 to length(Y) do 7: if I ≥ 3 && Y(i) > initial_value then 8: initial_slope_maxi = −2 × Y(I — 2)−Y(I — 1)+Y(I + 1)+2 × Y(I + 2) 9: count = count + i 10: break 11: end if 12: end for 13: for i:= count to length(Y) do 14: slope = −2 × Y(I — 2) — Y(I — 1) + Y(I + 1) + 2 × Y(I + 2) 15: if slope > initial_slope_maxi then 16: en = 1 17: end if 18: if en == 1 then 19: if slope < initial_slope_maxi then 20: initial_slope_maxi = old_slope 21: break 22: end if 23: end if 24: end for 25: maxi = initial_slope_maxi As the initial value of the EMG signal was not zero, the initial maximum slope capture requirements were changed in this study by using the average of n data and multiplying it by reduction ratio λ as the initial_value. The calculation approach could be expressed as Equation (8) in which n is set as 1000 samples. Each sample was 3.3 ms, and 1000 pieces of data (about 3.3 s) were captured for analysis during the initial setting of the initial_value, so as to capture stable initial_value data. When the detected signal exceeds this initial_value, it means that muscles have contracted and there are waveform characteristics. After determining the initial_slope_maxi, the subsequent slope_threshold calculation and peak detection can be performed. n ( λ + 1) initial_value = × ∑ Y (n) (8) n 1 λ in the initial_value calculation is the gathered statistics parameter. In order to avoid the error signal exceeding the average initial value, the initial_value calculation was multiplied by λ + 1 to evade the error signal. As only the error signal needed to be prevented, the accuracy was enhanced and stabilized after increasing the initial_value by 1%. An excessive λ value could increase the initial_value and result in the failure to capture the peak and a reduction of accuracy. Λ was thus selected as 5% of the stable zone after the statistical analysis, as shown in Figure 9.
Electronics 2022, 11, x FOR PEER REVIEW 10 of 17 Electronics 2022, 11, x FOR PEER REVIEW 10 of 17 Electronics 2022, 11, 178 10 of 17 100 100 95 95 90 (%) (%) 90 Accuracy 85 Accuracy 85 80 80 75 75 70 0% 1% 2% 3% 4% 5% 6% 7% 8% 9% 10% 70 λ 0% 1% 2% 3% 4% 5% 6% 7% 8% 9% 10% Figure 9. λ statistical analysis. λ Figure 9. Figure statistical analysis. 9. λλ statistical analysis. Figure 10 illustrates a simulation of the peak detection by the ASM S & C algorithm, which employs Figure the initial_value 10 illustrates illustrates calculation topeak solvedetection the initialby detection error in this study. Figure 10 aasimulation simulation ofthe of the peak detection theASM by the ASM SS&&CCalgorithm, algorithm, The initial_slope_maxi which employs employs the was captured when initial_valuecalculation theinitial_value the calculationto detected tosolve solvethe signal theinitial exceeded initialdetection the detectionerror EMG errorininthis initial thisstudy. study. which value. The The ASM S & Cwas initial_slope_maxi captured 154 when captured peaks the successfully detected and there signal were eight exceeded the capture EMG er- initial The initial_slope_maxi was captured when the detected signal exceeded the EMG initial rors, value. leading The to an ASM S accuracy & C of 95.06%. captured 154 peaks successfully and there were eight capture errors, value. The ASM S & C captured 154 peaks successfully and there were eight capture er- leading rors, to antoaccuracy leading of 95.06%. an accuracy of 95.06%. 2 EMG value 2 EMG Peak 1.9 EMG value EMG Peak 1.9 1.8 (Vp-p) 1.8 1.7 (Vp-p) Voltage 1.7 1.6 Voltage 1.6 1.5 1.5 1.4 1.4 1.3 0 10 20 30 40 50 60 1.3 Time(s) 0 10 20 30 40 50 60 Figure Figure 10. Peak Peak detection detection of of the the ASM & CTime(s) ASM SS & algorithm. Figure 10. Peak detection of the ASM S & C algorithm. Table 3 compares the accuracy of the three algorithms. Equation (7) was used for the the calculation of accuracy. accuracy. While the S & C and ASM S & C algorithms could capture allthe While the S & C and ASM S & C algorithms could capture all of of Table 3 compares the accuracy of the three algorithms. Equation (7) was used for the correct peaks, incorrect the correct peaks, values incorrect werewere values captured as well. captured The ASM as well. S & CSalgorithm The ASM improved & C algorithm im- calculation of accuracy. While the S & C and ASM S & C algorithms could capture all of the accuracy proved of the S of the accuracy &C thealgorithm in capturing S & C algorithm the EMG’s in capturing theinitial EMG’s maximum slope, reduced initial maximum slope, the correct peaks, incorrect values were captured as well. The ASM S & C algorithm im- the subsequent reduced error rate,error the subsequent and resulted rate, andinresulted a final accuracy in a finalof accuracy 95.06%. Therefore, of 95.06%.the accuracy Therefore, proved the accuracy of the S & C algorithm in capturing the EMG’s initial maximum slope, of the the ASM Sof accuracy & the C method ASM S proposed & C method in this paper was proposed higher in this paper than wasboth the Sthan higher & Cboth and the ACD S reduced the subsequent error rate, and resulted in a final accuracy of 95.06%. Therefore, methods, by 20.98% and 23.71%, respectively. & C and ACD methods, by 20.98% and 23.71%, respectively. the accuracy of the ASM S & C method proposed in this paper was higher than both the S &Table 3. Algorithm C and accuracy ACD methods, bycomparison. 20.98% and 23.71%, respectively. Table 3. Algorithm accuracy comparison. TableAlgorithm 3. Algorithm SP (Data) Algorithm accuracy comparison. SP (Data) EP (Data) EP (Data) MP (Data) MP (Data) Accuracy (%) Accuracy (%) ACD ACD 132 132 31 31 22 22 71.35 71.35 Algorithm SP (Data) EP (Data) MP (Data) Accuracy (%) S &SC& C [22] [22] ACD 154 154 132 54 54 31 0 0 22 74.08 74.08 71.35 ASM ASM S& S & C 154 9 0 95.06 S& CC[22] 154 154 9 54 0 0 95.06 74.08 ASM S & C 154 9 0 95.06
Electronics 2022, 11, x FOR PEER REVIEW 11 of 17 Electronics 2022, 11, x FOR PEER REVIEW 11 of 17 Electronics 2022, 11, 178 11 of 17 4. Experimental Results 4.Electromyography 4.1. Experimental Results 4. Experimental Results 4.1.In 4.1. Electromyography muscle contractions, the signal generated by the potential difference between both Electromyography ends isIn called In musclethecontractions, muscle EMG signal.the contractions, Regarding the the non-invasive signalgenerated signal generated bythe by measurement thepotential potential method difference difference usedboth between between in both this paper, ends is disposable called the EMGelectrodes signal. were attached Regarding the to the skin non-invasive to observe measurement ends is called the EMG signal. Regarding the non-invasive measurement method used in muscle activities. method used in The thisexperiment thispaper, data paper,disposable were disposable the muscle electrodes electrodes states were were of attached attached making to the to the a fist skinskin once per second to observe to observe muscle within muscle one activities. activities. The minute, experimentand the The experiment dataexperiment data the were were platform the muscle muscle was states the of states of makingdesigned making sensing a fist oncea fist endper peronce secondof within this system. second within one Theone minute, disposable minute, electrodes and the were attached experiment to platform the flexor was thedigitorum designed superficialis sensing end and the experiment platform was the designed sensing end of this system. The disposable muscle of this (FDS) system. andThe the flexor digitorum disposable electrodes were profundus electrodes attachedwere to themuscle attached (FDP), flexortodigitorum as shown the flexor digitorumin Figure superficialis 11. (FDS) superficialis muscle muscle (FDS) and the and flexor the flexor digitorum profundus muscle (FDP), digitorum profundus muscle (FDP), as shown in Figure 11. as shown in Figure 11. Figure 11. Electrode position stereogram. Figure11. Figure 11.Electrode Electrodeposition positionstereogram. stereogram. The TriAnswer platform was strapped to the user’s wrist to capture EMG signals via a patch. The The The capturedplatform TriAnswer TriAnswer EMG platformsignals was were wasstrapped transmitted strapped toto the the to the user’s user’s cloud wrist wrist totovia the EMG capture capture Bluetooth EMG inter- signals signals viavia a face and a patch. patch. displayed TheThe capturedin captureda smartphone EMG EMG signals signals app. werewere transmitted transmitted to the to the cloudcloud via via the the Bluetooth Bluetooth inter- interface face and The andoriginalinsignal displayed displayed in aand the app. smartphone a smartphone filtered app.signal are shown in Figure 12. Apparently, the curve Thewas smoothened Theoriginal originalsignal signalafter andfiltering. and thefiltered the In Figure filtered 12, signalare signal EMG areshown shown signals are 12. inFigure in Figure obviously cleaner 12. Apparently, Apparently, the the after curvefiltration, curve was which could wassmoothened smoothened afterimprove after the Indetection filtering. filtering. 12,ability In Figure Figure 12, EMG EMG of the ASM signals signals S & obviously are C algorithm. are obviously cleaner cleaner after after filtration, filtration, whichwhich could could improve improve the detection the detection abilityability of the of ASMtheSASM&CS & C algorithm. algorithm. Voltage (Vp-p) 3 Raw data Voltage (Vp-p) Raw data 2 3 1 2 0 1 0 2 4 6 8 10 12 14 16 18 20 0 0 2 4 6 8 Time(sec) 10 12 14 16 18 20 Time(sec) Voltage (Vp-p) 3 After filter Voltage (Vp-p) After filter 2 3 1 2 0 1 0 2 4 6 8 10 12 14 16 18 20 0 0 2 4 6 8 Time(sec) 10 12 14 16 18 20 Time(sec) Figure Figure12. 12.EMG EMGfiltering filteringcross crossreference. reference. Figure 12. EMG filtering cross reference. Figure 13 shows the spectrum analysis of Figure 12. Figure 13a presents the spectrum distribution of the EMG signals captured. Figure 13b displays the spectrum distribution of
Electronics 2022, 11, x FOR PEER REVIEW 12 of 17 Electronics 2022, 11, x FOR PEER REVIEW 12 of 17 Electronics 2022, 11, 178 12 of 17 Figure 13 Figure 13 shows shows the the spectrum spectrum analysis analysis of of Figure Figure 12. 12. Figure Figure 13a 13a presents presents the the spectrum spectrum distribution of the EMG signals captured. Figure 13b displays the spectrum distribution distribution of the EMG signals captured. Figure 13b displays the spectrum distribution of the of the EMG EMG signals signals after after moving moving average. average. As As seen, seen, there there are are obviously obviously fewer fewer noises noises in in the the the EMG signals spectrum after below distribution moving60 average. Hz after As seen, the movingthere are obviously fewer noises in the average. spectrum distribution below 60 Hz after the moving average. spectrum distribution below 60 Hz after the moving average. Amplitude Amplitude Amplitude Amplitude Figure 13. Figure 13. Spectrum analysis analysis of the the EMG signals. signals. Figure 13. Spectrum Spectrum analysis of of the EMG EMG signals. (a) presents the spectrum distribution of the EMG signals captured. (b) displays the spectrum distribution of the EMG signals after moving average The proposed The proposed system system controls controls the the actions actions ofof the the VR VR games games by by making making aa fist.fist. When When an an EMG EMGThe peak peak is detected, is detected, proposed systemthe output the output is controlsisthe 1; otherwise, 1; otherwise, actions of thethe theVRpeak peak detection detection games is 0. The is 0. The by making experiment experiment a fist. When an data EMGare data are peaktheissame the same as the detected, as thethe data for the output data for the is 1;moving average. otherwise, moving The detection the peak average. The ASM SS & ASM & isC 0. C algorithm is used The experiment algorithm is used to to determine data determine the are the same initial_value first, as the data first, the initial_value for the which moving which is then used average. is then used toto Thefind ASM find the initial_slope_maxi, theSinitial_slope_maxi, & C algorithm is as as usedwell wellto as as the determineslope_threshold. the slope_threshold. Whenever the initial_value two first, which Whenever consecutive is then usedEMG two consecutive EMG to findslopes exceed the slope_threshold, the initial_slope_maxi, slopes as well as exceed the slope_threshold, the the peak can be be identified slope_threshold. the peak can identified Whenever and the and the maxi twomaxi can be consecutive can be updated. updated. EMG slopesThe peak The peak exceedpulse is generated based slope_threshold, the is pulse generated based the on onpeakthe peak, thecan and the be identified peak, and the VR VRandgame is the maxi game controlled by can be updated. is controlled the peak pulse. The pulse. by the peak Figure peak pulse Figure 14 shows is generated 14 shows the the basedASM ASM on SSthe & peak, & C algorithm C algorithm detecting and thedetecting VR gamethe the EMG signal is controlled EMG signal and by and generating the peak the peak pulse. the generating peak Figure pulse. 14pulse. showsThe The blue line the blue ASM line is S & isC the EMG algorithm value, the detecting pink the dotted EMG line signal is the and initial value generating of the the EMG, peak and pulse. the EMG value, the pink dotted line is the initial value of the EMG, and the red dotted line the The red blue dotted line isline the is is the peak EMG the peak pulse value,pulse generated the pink after dottedafter generated line isthethe the peak peak is detected. initial is detected. TheEMG, value of the The peak pulse peak pulse and the is used is useddotted red for control- for control- line is ling the the peak actions pulse of the generated VR ling the actions of the VR figure. figure. after the peak is detected. The peak pulse is used for controlling the actions of the VR figure. EMG value 3.5 EMG value 3.5 initial value initial value peak pulse peak pulse 3 3 2.5 (Vp-p) 2.5 Voltage(Vp-p) 2 2 Voltage 1.5 1.5 1 1 0.5 0.5 0 00 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20 Time(s) Time(s) Figure 14. Figure 14. Figure Peak pulse Peak pulse 14. Peak generated pulse generated with generated with the with the EMG the EMG signal. EMG signal. signal. 4.2. Virtual Reality VR can make the users feel like they are at the scene, making it feasible for training and fitness uses. In this study, VR was combined with health care to make health management more interesting. This system, which employs an HTC VIVE VR headset, differs from other systems that it uses EMG for game control. As shown in Figure 15, the serial port needs to be selected to connect with the control terminal before the game could start. When the
4.2. Virtual Reality VR can make the users feel like they are at the scene, making it feasible for training and fitness uses. In this study, VR was combined with health care to make health man- Electronics 2022, 11, 178 agement more interesting. This system, which employs an HTC VIVE VR headset, 13 differs of 17 from other systems that it uses EMG for game control. As shown in Figure 15, the serial port needs to be selected to connect with the control terminal before the game could start. Whenstarts, game the game starts, theisdirection the direction is by controlled controlled by the the left and lefthands right and right hands to evade to evadeand obstacles, ob- stacles, and the speed increases as the game progresses. The game ends when the speed increases as the game progresses. The game ends when the player collides with the player collides an with obstacle. Theanfront-viewing obstacle. The angle front-viewing angle displays the displays current timetheandcurrent the besttime and time. Thethe best game time. The picture game picture is shown in Figureis shown 16. in Figure 16. Figure15. Figure 15.Game Gamesetting. setting. Figure16. Figure 16.Game Gamepicture. picture. Figure17 Figure 17isisthe theVR VRsystem systemflowchart. flowchart.The TheBluetooth BluetoothCOMCOMportportneeds needsto tobe beconnected connected atthe at thebeginning beginning of of the the game. game. When When the the game game starts starts running, running, the the EMG EMG state state is is detected. detected. Whenthe When theleft-hand left-handEMG EMGsignal signaltrigger triggerisisdetected, detected,the thegame game figure figurewould wouldwalk walk toto the the left; left; whenthe when theright-hand right-handEMG EMG signal signal trigger trigger is is detected, detected, the the figure figure would would dodge dodge toto the the right. right. Thescreen The screenisisupdated updatedperiodically. periodically.When Whenthethegame gamefigure figureencounters encountersananobstacle, obstacle,whether whether the playtime exceeds the record is determined; if yes, the best time would be updated and displayed; if not, the best time would remain and the game would end.
Electronics 2022, 11, x FOR PEER REVIEW 14 of 17 Electronics 2022, 11, 178 14 of 17 the playtime exceeds the record is determined; if yes, the best time would be updated and displayed; if not, the best time would remain and the game would end. No No No Update Print Time Yes Collision Yes Start Comport Instantiate Detection Time END Yes Save and Time Print Time Figure 17. VR system flowchart. Figure 17. VR system flowchart. 4.3. The 4.3. App The App The proposed The proposed system systememploys employsMITMITApp AppInventor Inventor to to design designa mobile a mobile appapp for conven- for con- ient observation venient observation andandrecording of the recording user.user. of the The The app app communicates communicates withwiththe sensing end the sensing via Bluetooth, end and is via Bluetooth, andworkable without is workable beingbeing without connected to a VR connected to headset. The start a VR headset. Themenu start displays menu the connection, displays settings, the connection, and related settings, app information and related (see Figure app information 18A). The (see Figure user’s 18A). The name, name, user’s gender,gender, and age areage and shown on the on are shown connection page. The the connection user page. Thecould userchoose to con- could choose nect to the smartphone connect the smartphoneto either to both eitherhands both or one or hands hand. oneWhen hand.the Whencorresponding sensing the corresponding end is connected, sensing the current end is connected, EMG diagram the current EMG diagramis displayed instantly, is displayed so that instantly, the users so that can the users check can theirtheir check muscle conditions. muscle Figure conditions. 18B shows Figure the app’s 18B shows state state the app’s after the afterright-hand sens- the right-hand sensing ing end isend is connected. connected. Thecould The data data could be savedbe saved duringduring the connection, the connection, and theand EMGthesignals EMG signals could be could be automatically automatically saved saved in the in the micro micro database database by clicking by clicking the save the save button. button. On On the the otherother hand,hand, datadata saving saving couldcould be stopped be stopped immediately immediately by clicking by clicking the the stopstop button button or or the the “Connection Off” button. The previously-accessed EMG “Connection Off” button. The previously-accessed EMG data could be queried ondata could be queried on the “View “View Data” Data” page. page. Figure Figure 18C 18C shows shows thethe EMG EMG data data found found using using “View “View Data”. Data”. The user information could be modified in the setting interface (Figure 18D). In addition to EMG signals, the app could collect biomedical data (ECG, EEG, and blood oxygen), making it highly highly suitable suitable for for long-term long-term collection collection and and observation. observation. On the menu displayed when the users open the app, the main operation includes the “Connect” and “Set” options. “View Data” or “Scan” could be selected in the connect setting. “View Data” is for EMG signal record queries, by which the saved EMG signal data could be observed. “Scan” is to view the list of nearby Bluetooth devices, through which the wearable system with intelligent and wireless features could be selected and the corresponding Bluetooth connection could be established. The current physiological condition could be observed after the connection is verified, the data could be saved by clicking “Save” while in the connected state, and the Bluetooth connection could be disconnected by clicking “Disconnect”. The user data could be input in the Set part, and “Enter” could then be chosen to complete the setting, or the user could choose “Clear” to clear the user data. The operation response time of this system is shown in Table 4. There are 320 pixels in one EMG image and it takes 0.1 s for one pixel, so totaling 30 s to draw the whole EMG image and the sample rate of the hardware capturing EMG signals is 300 Hz. Table 4. The system operation response time. Event Time EMG image update 30 (s) Hardware sampling 3.3 (ms) (A) (B)
could be automatically saved in the micro database by clicking the save button. On the other hand, data saving could be stopped immediately by clicking the stop button or the “Connection Off” button. The previously-accessed EMG data could be queried on the “View Data” page. Figure 18C shows the EMG data found using “View Data”. The user information could be modified in the setting interface (Figure 18D). In addition to EMG Electronics 2022, 11, 178 signals, the app could collect biomedical data (ECG, EEG, and blood oxygen), making it 15 of 17 highly suitable for long-term collection and observation. Electronics 2022, 11, x FOR PEER REVIEW 15 of 17 (A) (B) (C) (D) Figure 18. Mobile Figure 18. app usage Mobile apppictures. (A) Home(A) usage pictures. page, (B) page, Home Connection interface, (C) (B) Connection Reading interface, (C)the Reading the EMG data, EMG(D) User(D) data, data settings. User data settings. On5.the Conclusions menu displayed when the users open the app, the main operation includes the “Connect” and This “Set” study options. an developed “View Data” wireless intelligent or “Scan” could besystem wearable selected in the connect in combination with IoT setting.and “View Data” is forgames VR interactive EMGinsignal orderrecord queries, to allow bymonitor users to which the theirsaved EMG signal physiological conditions data could be observed. anytime “Scan” is and through wearable to view the list ofThis IoT devices. nearby Bluetooth system combined devices, EMGthrough with VR, so as to enhance users’ enjoyment during exercise and rehabilitation. The EMG which the wearable system with intelligent and wireless features could be selected and peak detection accuracy of the the corresponding ASM Sconnection Bluetooth & C algorithm could proposed in this paper be established. was higher The current than that of the physiological S &could condition C andbeACD algorithms observed by connection after the 20.98% andis23.71%, respectively. verified, A mobile the data could appby be saved was also clicking “Save” while in the connected state, and the Bluetooth connection could be dis- connected by clicking “Disconnect”. The user data could be input in the Set part, and “En- ter” could then be chosen to complete the setting, or the user could choose “Clear” to clear the user data.
You can also read
