Real-Time Gait Cycle Parameter Recognition Using a Wearable Accelerometry System

Transcription

1 Sensors 2011, 11, ; doi: /s OPEN ACCESS sensors ISSN Article Real-Tie Gait Cycle Paraeter Recognition Using a Wearable Acceleroetry Syste Che-Chang Yang 1,3, Yeh-Liang Hsu 1,3, *, Kao-Shang Shih 2 and Jun-Ming Lu Departent of Mechanical Engineering, Yuan Ze University, 135 Yuan-Tung Rd., Chung-Li, Tao-Yuan 320, Taiwan; ccyang@saturn.yzu.edu.tw Division of Orthopedics, Shin Kong Wu Ho-Su Meorial Hospital, 95 Wen Chang Rd., Shih Lin District, Taipei 111, Taiwan; scorelin@s28.hinet.net Gerontechnology Research Center, Yuan Ze University, 135 Yuan-Tung Rd., Chung-Li 320, Taiwan; jlu@saturn.yzu.edu.tw * Author to who correspondence should be addressed; ehsu@saturn.yzu.edu.tw; Tel.: ; Fax: Received: 13 June 2011; in revised for: 13 July 2011 / Accepted: 18 July 2011 / Published: 25 July 2011 Abstract: This paper presents the developent of a wearable acceleroetry syste for real-tie gait cycle paraeter recognition. Using a tri-axial acceleroeter, the wearable otion detector is a single waist-ounted device to easure trunk accelerations during walking. Several gait cycle paraeters, including cadence, step regularity, stride regularity and step syetry can be estiated in real-tie by using autocorrelation procedure. For validation purposes, five Parkinson s disease (PD) patients and five young healthy adults were recruited in an experient. The gait cycle paraeters aong the two subject groups of different obility can be quantified and distinguished by the syste. Practical considerations and liitations for ipleenting the autocorrelation procedure in such a real-tie syste are also discussed. This study can be extended to the future attepts in real-tie detection of disabling gaits, such as festinating or freezing of gait in PD patients. Abulatory rehabilitation, gait assessent and personal telecare for people with gait disorders are also possible applications. Keywords: acceleroetry; acceleroeter; Parkinson s disease; gait; obility

2 Sensors 2011, Introduction Gait dynaics reflect one s obility which can be affected by physical ipairent, age progress and changes in health status. Abulatory gait paraeters can be iportant easures to assess functional ability, balance control and to predict fall risk. Individuals with degenerative obility, e.g., Parkinson s disease (PD) patients or older adults usually have gait disorders such as reduced walking speeds with increased cadences, reduced step/stride lengths, and increased inter-stride variability [1]. PD patients of advanced stage ight have encountered episodic gait disturbances, like festinating or even freezing of gaits (FOG) that could lead to falling and adverse health outcoes [2,3]. Regularity, rhyth and syetry are iportant gait cycle paraeters that can be apparently altered in walking patterns aong people of varied obility [2-4]. Therefore, the onitoring of the above gait cycle paraeters can be beneficial to assess the obility and risk of occurrence of episodic gait disturbances. Gait evaluation is frequently based on observational interpretations which are subjective and ay vary aong clinicians or investigators. As a consequence, gait onitoring and analysis techniques have been widely developed and studied. Gait dynaics can be accurately easured by using optical otion capture systes which utilize high-speed infrared caeras to record the three-diensional positions of retro-reflective arkers attached to the joints and segents of the huan body [5]. Gait detection techniques utilizing pressure sensors ebedded in an overground walkway [6] have also been used. These techniques can detect foot contact (heel strike and toe-off) and even the foot pressure distribution to investigate teporal gait paraeters. However, those systes are expensive, and the sophisticated instruentation requires specialized personnel. Therefore the uses of those systes are usually liited in laboratory or clinical environents. Sipler systes based on pressure detection, such as the portable in-shoe pressure easureent syste have also been presented [7,8]. The systes utilizing in-shoe pressure detection can only provide siple teporal gait easures while acceleroeter-based or video-based systes can provide teporal and spatial gait easures, even accurate easureent of lower libs and body oveent. Acceleroetry using wearable systes has drawn a vast aount of research interest in the study of huan oveent. Acceleroeters have widely been used in wearable systes for oveent classification, fall detection, estiation of energy expenditure and gait analysis [9,10]. Acceleroeters in cobination with gyroscopes that easure angular velocity and accurate orientation have also been developed [11,12]. Though the pathological gaits have been well studied and described, only a few studies have investigated recognition of abnoral gaits using wearable acceleroetry systes. A shank-ounted acceleroeter was used to onitor the FOG in PD patients by eans of a freeze index coputed by frequency spectral analysis. However, the power spectral analysis can hardly be perfored in real-tie on copact wearable systes [13]. A wearable syste using ARM7 processor was also deonstrated to detect FOG in real-tie fro every collected 0.32 s acceleration data [14]. Due to the coputation constraints, it was reported that a longer saple data will produce longer latency of the syste which ight not be acceptable for practical uses. The developent of cost-effective approaches to real-tie gait onitoring is iportant and beneficial. This paper presents the developent of a wearable acceleroetry syste for real-tie gait cycle paraeter recognition. A waist-ounted wearable otion detector was designed to easure trunk accelerations during walking. The autocorrelation procedure is ipleented in the syste to

3 Sensors 2011, derive several gait cycle paraeters, including cadence, step regularity, stride regularly and step syetry. For validation purposes, five PD patients and five young healthy adults were recruited in order to investigate whether the gait cycle paraeters aong the two subject groups of different obility can be quantified and distinguished by the proposed syste. Practical considerations and liitations for ipleenting real-tie gait cycle paraeter recognition in an ebedded wearable syste are also discussed. If the gait cycle paraeters can be identified in real tie, continuous detection of gait variability would be possible. This paper can lead to a future developent of wearable systes enabling real-tie recognition of abnoral gaits, such as shuffling, festinating, or freeze of gait in PD patients, by exaining patterns of various gait cycle paraeters and other possible characteristics in acceleration patterns. Abulatory rehabilitation, gait assessent and personal telecare for people with gait disorders are also possible applications. 2. Method 2.1. Instruentation The wearable otion detector is a single waist-ounted device that easures trunk accelerations of huan oveents. Figure 1 shows the circuit board assebly and the prototype of the wearable otion detector. It uses a tri-axial acceleroeter odule (KXPA4-2050, Kionix) that senses accelerations in the sensitivity of 660 V/g over the selected range of ±2 g. The acceleroeter odule has an internal built-in low pass filter at cut-off frequency of 50 Hz. This circuit liits the bandwidth of the signal outputs and therefore reduces the higher frequency coponents which are not related to actual huan oveents. A PIC icrocontroller (PIC18LF6722, Microchip) offers flash eory of 128 kbytes and SRAM of 3,936 bytes. It saples the analog output signals via a 10-bit A/D conversion at the sapling rate of 50 Hz. Real-tie signal processing can be ipleented in the PIC icrocontroller. The wearable otion detector also uses a wireless 2.4 GHz ZigBee RF odule (XBee 2.0, Digi International) to transit real-tie recognized gait cycle paraeters to a personal coputer (PC). The gait cycle paraeters can be displayed on the PC screen for teleonitoring and data logging. Powered by 3 AAA batteries (DC4.5V), the wearable otion detector easures in size and 120 g in weight. The battery life can last up to approxiately 50 h when the wearable otion detector is continuously in use. Figure 1. The prototype of the wearable otion detector.

4 Sensors 2011, The wearable otion detector was clipped to the pant belt of the subjects. The detector is positioned near the iddle between the anterior superior iliac spine and the right iliac crest around the pant belt. The pant belt on the subject was fastened ediu to tight without causing discofort to the subject, as this adjustent can reduce isalignent of orientation and vibration of the instruent which will generate signal artifacts and noises Subjects and Gait Data Collection For validation purposes, five elderly Parkinson s disease patients (four ales and one feale, 78 ± 9.8 yr) diagnosed as Hoehn & Yahr (H&Y) stage II to III and five young healthy subjects without obility ipairent (all ales, 26 ± 3.1 yr) were recruited for gait data collection. The data collection was approved by the Institutional Review Board (IRB) at the Far-Eastern Meorial Hospital, Taiwan. The recruited subjects were provided with necessary inforation about the easureent and they gave their infored consents before the data collection. Before gait data easureent, the Tied Up and Go (TUG) test, which is a validated siple and quick easure for obility assessent [15], was conducted to quickly screen the obility level of all the subjects. The tie for the PD patients to perfor the TUG test is 23.9 ± 7.9 s, while the young healthy subjects took only 10.6 ± 2.2 s. The distinct difference in the TUG test results show a generally degenerative obility level in the PD patient group. The 5-eter-walk test (5WMT) was conducted in a laboratory. In the 5MWT, the subjects wore the wearable otion detector at their waists while walking on a 5-eter level walkway three ties at their own noral and faster walking paces. The accelerations along the vertical (VT), antero-posterior (AP) and edio-lateral (ML) directions were recorded at the sapling rate of 50 Hz. The initiation of data sapling of the wearable otion detector triggered the start of synchronized video recording during the test for gait observation and cadence validation Gait Cycle Paraeters Recognition Walking can be generally regarded as a repeated oveent of huan body. Therefore the easured accelerations during walking should also reveal periodic signal patterns. The autocorrelation procedure is a ethod to estiate the repeating characteristics over a signal sequence containing periodic patterns and irregular noises. Moe-Nilsson et al. have deonstrated the fundaentals of the autocorrelation procedure for coputing gait cycle paraeters, which is the basis of the recognition ethod in this study [16]. In the work by Moe-Nilsson et al., and the subsequent work by Yang et al. [17] and Keenan et al. [18] using autocorrelation for gait cycle analysis, the gait cycle paraeters were coputed in an off-line anner. In this paper, the autocorrelation procedure is ipleented in an ebedded wearable syste for real-tie gait cycle paraeter recognition, which extends its possible applications. The autocorrelation procedure for coputing gait cycle paraeters is described below. Practical considerations and liitations for ipleenting the autocorrelation procedure in the ebedded syste for real tie gait cycle paraeter recognition are also discussed. Consider a tie-discrete acceleration sequence containing N signal points [ x 1, x2, x3,, x N 1, xn ], Equation (1) calculates the autocorrelation coefficient a, which is the su of the products of x i

5 Sensors 2011, ultiplied by another signal x i + at the given phase shift. The phase shift can be either positive or negative integers fro 0 to N 1, or fro 0 to 1 N. Therefore, fro an N-point acceleration sequence, its autocorrelation sequence A = [ a, a + 1,, a0, a1, a 1, a] can be represented by 2N 1 autocorrelation coefficients obtained at every phase shift. The autocorrelation sequence can either be biased or unbiased. The unbiased autocorrelation sequence as shown in Equation (2) is preferred because the biased ethod generates noticeable attenuation of coefficient values next to the zero phase shift fro a liited nuber of data: N a = xi xi+ i= 1 N a unbiased 1 = xi xi N + i= 1 (1) (2) The segents fro a to a 1 and fro a 1 to a in an autocorrelation sequence are syetric with its zero phase shift a 0 located at the center of the sequence. Noralized to 1 at the zero phase shift a 0, only the right half segent a 0 to a of the autocorrelation sequence is considered for siplicity. Figure 2 depicts an exaple of an autocorrelation sequence coputed fro the VT accelerations easured at waist during noral walking paces. The first coefficient peak D 1 next to the zero phase shift indicates the first doinant period, and the second peak D 2 the second doinant period. The peaks D 1 and D 2 can be detected by a siple derivative-based ethod and zero-crossing identification, which are coonly used in detecting peaks in physiologic signals, such as PQST points in ECG signals. The two peaks can also be found on the autocorrelation sequence coputed fro the AP acceleration sequence. Fro repeated observations, the two peaks on the VT autocorrelation sequence appear in the sae positions as the peaks on the AP autocorrelation sequence. Therefore superiposing the VT and AP autocorrelation sequences can better highlight the exact positions of the two peaks. Figure 2. The exaple of an autocorrelation sequence coputed fro the vertical acceleration easured at waist during walking.

6 Sensors 2011, The following gait cycle paraeters can be derived fro the autocorrelation sequence: Step regularity and stride regularity: A signal sequence with perfectly repetitive pattern produces its autocorrelation sequence containing the peak agnitudes identical to its zero phase shift at every doinant period, i.e., the agnitudes at every doinant period is 1 for a noralized autocorrelation sequence. The agnitude D 1 represents step regularity, as defined by Moe-Nilssen et al. [16]. This is because the first doinant period indicates the axial siilarity between the acceleration sequence and its -point shifted duplicate. The -point span approxiates the duration of a step. Siilarly, the second doinant period indicates the axial siilarity between the acceleration sequence and its 2-step shifted sequence, and therefore the agnitude D 2 represents stride regularity [16]. Note that the first and second doinant periods do not represent which of the steps (left-leg or right-leg) as there is no such inforation given in the autocorrelation procedure. Step syetry: Step syetry is defined as the ratio of step regularity to stride regularity, or D / D 1 2 that indicates the syetry between two steps of both legs [16]. In this paper, the step syetry is D / D if 1 2 D2 D1, and D / D when 2 1 D 2 < D1. Note that this definition is altered fro the definition originally given by Moe-Nilssen et al. [16] and its odified version by Yang et al. [17]. In this definition, the step syetry always ranges fro 0 to 1, which would be ore interpretable. Cadence: Cadence is the step rate per inute. Let S be the nuber of steps taken over the tie period t (in second). Cadence can thus be expressed as Equation (3). The nuber of steps S can be expressed as the nuber of the total saples N divided by the nuber of the coefficients n between the zero phase shift and the first doinant period, i.e., S = N / n. The tie period t during walking can also be alternatively expressed as N divided by the sapling frequency f, i.e., t = N / f. As a result, cadence (c) can be estiated by Equation (4), which was given by Moe-Nilssen et al. [16]: S c = 60 (3) t N S f c = 60 = 60 n = 60 (4) t N n f A non-overlapping sliding window technique is used to cyclically produce gait cycle paraeters in real tie fro the VT, AP, and ML accelerations. Longer window lengths can produce ore precise gait cycle paraeters because the data of ore steps is included. However, considering the real-tie processing constraints, longer window lengths will cause longer coputation latency and less reserved argin of eory capacity. Taking the hardware eory capacity and coputation latency of the wearable otion detector into account, the window length is set at constant 3.5 s. According to the assued average cadence of 105 steps/in that approxiates regular walking cadence steps/in [19,20], the choice of 3.5-second data can includes approxiately 6 steps, which should be sufficient for autocorrelation processing.

7 Sensors 2011, Results and Discussion 3.1. Results of the Experient To derive the gait cycle paraeters, the VT acceleration was first used to copute the autocorrelation sequence because the VT coponent can represent the characteristics of each step for classification of walking [21]. In this study, the autocorrelation sequences coputed fro the healthy young subjects exhibit a ore sooth and onotonic pattern, while the counterparts obtained fro the PD patients contain visible fluctuations and are less regular. Figure 3 shows one exaple of such observations in this study. The peak agnitudes at the doinant periods fro a PD patient s pattern are relatively lower than that fro a healthy young subject s pattern. Figure 3. The exaple of autocorrelation sequences (VT acceleration) coputed fro a young healthy subject (above) and a PD patient (below). Figure 4 shows the autocorrelation sequences coputed fro VT and AP accelerations obtained fro a healthy subject and a PD patient. The young healthy subject has better step and stride regularity (larger agnitudes at the peaks D 1 and D 2 ) than the PD patient. This shows that the periodic characteristics of gait in the PD patient are not steadily and perfectly reproduced. It is shown that the doinant periods on the VT and AP patterns coincide with each other, even though the two patterns ay vary differently. The coparison of both the VT and AP autocorrelation sequences can iprove the accuracy in identifying the doinant periods when the doinant periods cannot be clearly deterined fro the VT autocorrelation sequence alone. Table 1 shows the ean values of the recognized gait cycle paraeters of all subjects of the two groups. Coparing the cadences derived fro the autocorrelation procedure, the average cadence of the PD group (102.2 ± 15.2 steps/in) is slightly higher than that of the healthy group (98.6 ± 5.8 steps/in). The elevated cadences at longer TUG tie for the PD patient group indicate reduced step length and walking speed, which confors to the literature result [1]. In fast 5MWT, the average cadence of the PD group was ± 15.6 steps/in, which was approxiately 5.8% increased fro their noral cadences. The healthy group had an average cadence of ± 6.2 steps/in in fast 5MWT, which was approxiately 16.9% increased fro their noral cadence. This indicates a liited perforance argin for the PD group due to their degenerative obility.

8 Sensors 2011, For the gait regularity and syetry, the PD patient group has the step regularity of 0.39 ± 0.16 and the stride regularity of 0.43 ± 0.2 during noral walking paces. The healthy group has higher step regularity (0.61 ± 0.14) and stride regularity (0.79 ± 0.09). Siilar trends can be observed in their fast walking paces. Therefore, the result shows that the PD patients cannot well regulate their repeating steps and strides copared with the healthy group. Note that the step syetry during noral walking fro the PD group is slightly higher than that fro the healthy group, while the step syetry during fast walking in the PD group is lower than that in the healthy group. This ixed results regarding step syetry need further investigations. In the current definition of step syetry, it is possible to obtain higher step syetry fro the gaits with both low step regularity and stride regularity that are close to each other. Altered definitions of step syetry have been used [18], and it is also of iportant interest in the future to investigate which definition can be suitably applicable. Figure 4. The exaple of the VT and AP autocorrelation sequences coputed fro a young healthy subject (above) and a PD patient (below).

9 Sensors 2011, Table 1. Gait cycle paraeters derived fro the subjects. 5MWT (noral) 5MWT (fast) PD Healthy Mean ± SD p-value ean ± SD p-value TUG test 23.9 ± 7.9 s ± 2.2 s Cadence ± ± Step regularity 0.39 ± ± Stride regularity 0.43 ± ± Step syetry 0.81 ± ± Cadence ± 15.6 n/a ± Step regularity 0.37 ± 0.17 n/a 0.76 ± Stride regularity 0.47 ± 0.12 n/a 0.80 ± step syetry 0.75 ± 0.22 n/a 088 ± The cadence obtained by the ethod was copared with that easured fro the synchronized video. The ean absolute percentage error is 4.89%. As the subjects repeated each test ite three ties during the data collection, it is hypothesized that the subjects repeated the tests without significant variance so that the results of gait cycle paraeter recognition can be valid. Therefore, the one-way ANOVA test was used to investigate whether each test ite shows significant variance in both the subject group. In general, there is no significant difference (significant level 0.05) between the results fro each test of the subjects. Note that because soe of the PD subjects were unable to perfor fast 5MWT, the test in the PD subjects fast 5MWT are excluded in Table 1 due to the liited data saples Discussion Several studies have reported the use of the vertical accelerations for autocorrelation procedures [16,17]. In this study the VT, AP and ML acceleration coponents were copared to exaine which axis is ost sensitive to steps and produces identifiable pattern related to the gait cycle paraeters. Fro our observation fro the autocorrelation sequences of the 10 test subjects, the ML acceleration coponent is considered least descriptive and least sensitive to walking oveent. This was also observed in a previous study by the authors [22]. In the study by Keenan et al. [18], the VT and AP coponents were also used for autocorrelation process for the sae reason even triaxial accelerations were easured. Though the ML acceleration coponent easured fro the back over the L3 region has been shown good results in autocorrelation procedure [16], this could be because of different positions to attach the devices were used. In the future developents, the ML accelerations ay be used to distinguish left leg or right leg steps [23]. The real-tie 3D orientation of the trunk can be synchronously coputed fro the vertical acceleration, or fro the cobination of the three acceleration coponents for better accuracy. However, trunk orientation reveals less inforation in interpreting gait patterns, and thus it is not used throughout the recognition procedure. The wearable otion detector developed in this study easures walking oveent at the sapling rate of 50 Hz. As a low-pass filter circuit (fc = 50 Hz) was internally built in the acceleroeter odule to reduce the signal coponents of higher frequencies that would not be relevant to huan oveents in daily living, ideally the sapling frequency above 100 Hz should be better. However, the sapling

10 Sensors 2011, rate 50 Hz here is used because of the liitations of available eory capacity (data to be buffered), coputation capability of the low-cost PIC icrocontroller, and the constraint of a real-tie syste requiring iniized cyclic processing tie. Moreover, intense and faster oveents could rarely occur in daily living hoe environents. Copounding the above considerations, the sapling rate 50Hz was used, though higher sapling rate is certainly better if capable hardware is available. Step regularity, stride regularity and step syetry obtained fro the autocorrelation procedure are the general teporal gait estiates in ters of signal periodic characteristics. Cadence can be easily validated by coparing the synchronized video recording. Tura et al. have validated the step and stride regularities to well correlate with the indices obtained fro step and stride duration easured fro pressure insoles [24]. Offline analysis using autocorrelation ethod usually processes large acceleration data to copute the gait cycle paraeters that indicate its overall gait perforances. Real-tie recognition of the gait cycle paraeters using cyclic processing ay be affected by a high instantaneous variability. An extra continuous 25-eter walk data fro three healthy young subjects in their constant walking paces was additionally used to check the effect of using shorter window lengths to calculate the gait cycle paraeters. As shown in Table 2, the gait cycle paraeters obtained fro the ultiple sliding windows (window length 3.5 s) can oderately to highly approxiate the gait cycle paraeters obtained fro the entire single window. However, for data in varied walking speeds, it was found that gait cycle paraeters obtained fro the ultiple sliding windows were sensitive and can reflect the episodic signal variability. Table 2. Gait cycle paraeters derived fro ultiple sliding windows and the entire single window. Cadence (steps/in) Mean coefficient of variance (CV) Multiple sliding windows Mean percentage error 1.21% 0.67% Step regularity 8.53% 2.44% Stride regularity 9.34% 4.47% Step syetry 7.78% 2.04% Mean ± SD Entire single window ± ± ± ± ± ± ± ± ± ± ± ± As real-tie gait cycle paraeter recognition is developed, continuous detection of disabling gaits could be applicable. Figure 5 shows the process flowchart of the algorith. After the data sapling (P1), the edian-filtering process (P2) is applied to the sapled data to eliinate signal spikes that are not related to huan oveent. The VT and AP autocorrelation sequences are generated (P3) fro the sapled data. In the process P4 both the VT and AP autocorrelation sequences are superiposed to obtain a peak-highlighted sequence which is used for peak detection. The P5 process locates the

11 Sensors 2011, positions of the first and second doinant periods in the VT-AP superiposed sequence. If the process D1 fails to identify the positions of the first and second doinant periods in the VT-AP superiposed sequence (D1 = No), the procedure returns to the process P1 to restart next data sapling. If that positions are identifiable (D1 = Yes), the peak search process (P6) finds the first and second doinant periods, D 1, and D 2 in the VT autocorrelation sequence according to the peak position located fro the VT-AP superiposed sequence. The gait cycle paraeters are then coputed in the process P7 and the estiates output is provided prior to the next new processes fro P1. With the available gait cycle paraeters in consecutive identification periods, a knowledge base of gait disorders and the pattern characteristics, i.e., shuffling, festinating and freeze of gaits can be integrated in the future to facilitate the potential capability in real-tie and continuous detection of disabling gaits in PD patients. Figure 5. The process flowchart of real-tie gait cycle paraeters recognition.

12 Sensors 2011, Conclusions This paper presents the use of a wearable otion detector for real-tie gait cycle paraeter recognition. The wearable otion detector is a single waist-ounted device that utilizes a tri-axial acceleroeter to easure trunk accelerations during walking. The autocorrelation procedure is used to estiate cadence, step regularity, stride regularity and step syetry fro the easured trunk accelerations in real tie. In this study, the gait cycle paraeters aong the two subject groups of different obility (PD patients and the young healthy adults) can be quantified and distinguished by the syste. The wearable otion detector has been developed by the authors for real-tie physical activity identification, and its connection to a telecare syste has also been presented [22]. The syste developed in this paper can lead to future research interests and developent regarding real-tie detection of gaits disorders in PD patients. Mobility assessent, abulatory rehabilitation for PD patients can be the possible applications. Acknowledgents This research is supported by the Division of Neurology and Orthopedics in Far Eastern Meorial Hospital, Taiwan. This support is gratefully acknowledged. References 1. Lowry, K.A.; Siley-Oyen, A.L.; Carrel, A.J.; Kerr, J.P. Walking stability using haronic ration in Parkinson s disease. Mov. Disord. 2009, 24, Hausdorff. J.M. Gait dynaics, fractals and falls: Finding eaning in the stride-to-stride fluctuations of huan walking. Hu. Mov. Sci. 2007, 26, Hausdorff, J.M. Gait dynaics in Parkinson s disease: Coon and distinct behavior aong stride length, gait variability, and fractal-like scaling. Chaos 2009, 19, Plotnik, M.; Hausdorff, J.M. The role of gait rhythicity and bilateral coordination of stepping in the pathophysiology of freezing of gait in Parkinson s disease. Mov. Disord. 2008, 23, S444-S Roiz, R.M.; Cacho, E.W.A.; Pazinatto, M.M.; Reis, J.G.; Cliquet, A., Jr.; Barasnevivius-Quagliato, E.M.A. Gait analysis coparing Parkinson s disease with healthy elderly subjects. Arq Neuropsiquiatr 2010, 68, Menz, H.B.; Latt, M.D.; Tiedeann, A.; Kwan, M.M.S.; Lord, S.R. Reliability of the GAITRite walkway syste for the quantification of teporal-spatial paraeters of gait in young and older people. Gait Posture 2004, 20, Hausdorff, J.M.; Ladin, Z.; Wei, J.Y. Footswitch syste for easureent of the teporal paraeters of gait. J. Bioech , Feery, V.; Moretto, P.; Renaut, H.; Thévenon, A.; Lensel, G. Measureent of plantar pressure distribution in heiplegic children: Changes to adaptive gait patterns in accordance with deficiency. Clin. Bioech. 2004, 17, Mathie, M.; Coster, A.C.F.; Lovell, N.H.; Celler, B.G. Acceleroetry: Providing an integrated, practical ethod for long-ter, abulatory onitoring of huan oveent. Physiol. Meas. 2004, 25, R1-R20.

13 Sensors 2011, Yang, C.C.; Hsu, Y.L. A review of acceleroetry-based wearable otion detector for physical activity onitoring. Sensors 2010, 10, Liu, T.; Inoue, Y.; Shibata, K. Developent of a wearable sensor syste for quantitative gait analysis. Measureent 2009, 42, Schepers, H.M; Asseldonk van, E.H.F.; Baten, C.T.M.; Veltink, P.H. Abulatory estiation of foot placeent during walking using inertial sensors. J. Bioech. 2010, 43, Moore, S.T.; MAcDougll, H.G.; Ondo, W.G. Abulatory onitoring of freezing of gait in Parkinson s disease. J. Neurosci. Meth. 2008, 167, Jovanov, E.; Wang, E.; Verhagen, L.; Fredrickson, M.; Fratangelo, R. Defog-A Real Tie Syste for Detection and Unfreezeing of Gait of Parkinson s Patients. In Proceedings of the 31st Annual International IEEE EMBS Conference, Minneapolis, MN, USA, 2 6 Septeber Podsiadlo, D.; Richardson, S. The tied Up & Go : A test of basic functional obility for frail elderly persons. J. A. Geriatr. Soc. 1991, 39, Moe-Nilssen, R.; Helbostad, J.L. Estiation of gait cycle characteristics by trunk acceleroetry. J. Bioech. 2004, 37, Yang, M.; Cheng, H.; Wang, H.; McLean, S.; Hall, J.; Harris, N. Assessing Acceleroeter Based Gait Features to Support Gait Analysis for People with Coplex Regional Pain Syndroe. In Proceedings of PETRA 10: the 3rd International Conference on Pervasive Technologies Related to Assistive Environents, Saos, Greece, June Keenan, D.B.; Wilhel, F.H. Classification of locootor activity by acceleration easureent: Validation in Parkinson disease. Bioed. Sci. Instru. 2005, 41, Hirasaki, E.; Moore, S.T.; Raphan, T.; Cohen, B. Effects of walking velocity on vertical head and body oveents during locootion. Exp. Brain Res. 1999, 127, Wang, N.; Redond, S.J.; Abikairajah, E.; Celler, B.G.; Lovell, N.H. Can triaxial acceleroetry accurately recognize inclined walking terrains? IEEE Trans. Bioed. Eng. 2010, 57, Karantonis, D.M.; Narayanan, M.R.; Lovell, N.H.; Celler, B.G. Ipleentation of a real-tie huan oveent classifier using a triaxial acceleroeter for abulatory onitoring. IEEE Trans. Inf. Technol. Bioed. 2006, 10, Yang, C.C.; Hsu, Y.L. Developent of a wearable otion detector for teleonitoring and real-tie identification of physical activity. Teleed. J. E Health 2009, 15, Zijlstra, W.; Hof, A.L. Assessent of spatio-teporal gait paraeters fro trunk accelerations during huan walking. Gait Posture 2003, 18, Tura, A.; Raggi, M.; Rocchi, L.; Cutti, A.; Chiari, L. Gait syetry and regularity in transfeoral aputees assessed by trunk accelerations. J. Neuroeng. Rehabil. 2010, 10, by the authors; licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the ters and conditions of the Creative Coons Attribution license (