//Logo Image
AuthorsChe-Chang Yang (2011-01-21); recommended: Yeh-Liang Hsu (2011-02-21).
Note: This article is the Chapter 6 of Che-Chang Yang’s doctoral thesis “Development of a Home Telehealth System for Telemonitoring Physical Activity and Mobility of the Elderly”.

Chapter 6. Gait cycle parameters recognition using the wearable motion detector

This chapter presents the use of the wearable motion detector described in the previous chapter and the autocorrelation procedure to recognize gait cycle parameters of Parkinson’s disease (PD) patients in real-time. The principles of the autocorrelation procedure dealing with accelerometry data are introduced. The gait cycle parameters derived from the measured accelerometry data of 5 elderly PD patients and 5 young healthy subjects are compared. The determinable and discriminative characteristics of the selected gait cycle parameters are highlighted. The possibility of developing a wearable system utilizing autocorrelation procedure to recognize abnormal gaits, such as shuffling, festinating is discussed.

6.1 Gait dynamics and measurements for Parkinson’s disease patients

Gait dynamics reflect one’s mobility which can be affected by physical impairment, age progress and changes in health status. Gait parameters extracted from complex ambulation dynamics can be important measures to assess functional ability, balance control and to predict risk of falling. Individuals with Parkinson’s disease (PD) suffer progressive motor impairments which can be characterized by resting tremor, bradykinesia, rigidity, and postural instability. PD results from deficiency of dopamine production due to the degenerative disorder of neurological functions. Levodopa (L-dopa)/Carbidopa treatment is the current therapy to reduce the PD symptoms. The Unified Parkinson’s Disease Rating Scale (UPDRS) [Fahn et al., 1987] and Hoehn and Yahr (H&Y) Modified Scale [Hoehn et al., 1967; Goetz et al., 2004] are the two major clinical measures to assess the stages of PD advances. In addition, the timed up-and-go test (TUG) [Podsiadlo et al, 1991] and the Berg Balance Scale (BBS) [Berg et al., 1989] are also the two assessment tools in terms of mobility.

PD affects gait disorders in the lower extremities, such as reduced walking speed with increased cadence, reduced step-length, and increased stride-to-stride variability [Lowry et al., 2008]. Shuffling gait is also commonly observed from moderate PD patients. The advanced PD patients may have experienced the episodic gait disorders, such as festination, hesitation and freeze of gait (FOG) that occur occasionally and intermittently and may lead to falling and adverse health outcomes (e.g., hip fracture) [Hausdorff, 2009].

Gait evaluation is frequently based on observational interpretation that may, however, vary among clinicians or investigators. As a consequence, monitoring and analysis techniques for quantitatively investigating the Parkinsonian and pathological gaits have been widely developed and studied to provide objective measures. Gait dynamics can be accurately measured by using the optical motion capture systems [Melo Roiz et al., 2010]. As shown in Figure 6.1, these optical systems use high-speed infrared cameras to record the three-dimensional positions of retro-reflective markers attached to the joints and segments of the human body during motion. The spatial-temporal gait parameters including velocity, stride length, cycle time and stance time can be identified from the kinematics data.

Figure 6.1 Typical attachment of the retro-reflective markers [de Melo Roiz et al., 2010]

The gait detection techniques utilizing pressure sensors embedded in an overground walkway [Menz et al., 2004] or portable in-shoe pressure measurement system have also been used [Femery et al., 2004]. These techniques detect foot contact (heel strike and toe-off) and even the foot pressure distribution to investigate spatial-temporal gait parameters. The walking speed, cadence, right/left step lengths, and even the toe-in/out angles can be captured.

Both gait monitoring approaches using optical techniques or pressure sensing techniques have to be limited in clinical/laboratory settings due to their expensive instruments, sophisticated system setup and the needs of specialized personnel.

Accelerometry measurement using wearable systems has drawn a vast amount of research interests in the study of human movement. It is only recently that a few numbers of studies have reported gait analysis using accelerometers while the accelerometer-based trials in movement classification, energy expenditure and fall detection have been largely studied [Mathie et al., 2004, Yang et al., 2010]. Though the Parkinsonian gaits have been well studied and described, only a few studies have investigated recognizing abnormal gaits based on wearable systems. A shank-mounted accelerometer was used to monitor the freezing of gait of the PD patients [Moore et al., 2008]. A frequency spectra analysis was used to compute the frequency components of gait data. A freeze index is defined as the ratio of two spectral bands of different frequency components 0.5-3Hz and 3-8Hz. However, the power spectral analysis cannot be performed in real-time on the compact wearable systems. A wearable system using ARM7 processor was also demonstrated to detect FOG in real-time from every collected 0.32s acceleration data [Jovanov et al., 2009]. Due to the computation constraints, it was reported that a longer sample data will produce longer latency of the system which might not be acceptable for a practical scenario.

This chapter presents the use of the wearable motion detector described in the previous chapter and the autocorrelation procedure to recognize gait cycle parameters of Parkinson’s disease (PD) patients in real-time. This chapter first demonstrates the extraction of gait cycle parameters from trunk accelerometry measurement. Several gait cycle parameters, including cadence, step regularity, stride regularly and step symmetry, can be derived from the acceleration signals in real-time using the autocorrelation procedure which can be implemented in microprocessor-based devices of limited computation capacity. Five PD patients and 5 young healthy subjects were recruited in a data collection session. The trunk accelerations during their guided walks were recorded using the wearable motion detector described in the previous chapter. The differences in the gait cycle parameters between PD patients and the healthy subjects were compared and discussed. The study in this chapter can lead to a future development of a wearable system for recognizing PD-related abnormal gaits, such as shuffling, festinating, or freeze of gait and falls in real-time, which can be important and beneficial in PD ambulation rehabilitation and personal tele-care applications.

6.2 Signal processing for extracting gait cycle parameters

This section introduces the autocorrelation procedure to compute temporal gait cycle parameters from acceleration signals measured by the wearable motion detector. The method shows different gait characteristics computed from the example of different walking patterns in terms of the cadence, step regularity, stride regularity, step symmetry.

6.2.1 Fundamentals of autocorrelation

Autocorrelation have widely been used in many engineering and scientific fields. Based on the Pearson product moment correlation, autocorrelation is a numerical method to calculate correlation of a signal with itself. It can be useful for analyzing spatial and temporal relationships between time-varying signals such as physiological behaviors and human movements [Nelson-Wong et al, 2009]. The use of autocorrelation in human movement research is not as common as in the other application fields [Moe-Nilssen et al., 2004; Keenan et al., 2005; Yang et al., 2010].

Autocorrelation is a method to estimate the repeating characteristics over a signal sequence containing periodic patterns and irregular noises. Consider a time-discrete signal sequence containing N signal points [,, , …, ], the Equation (6-1) calculates the autocorrelation coefficient , which is the sum of the products of xi multiplied by another signal  at the given phase shift m. The phase shift m can be either positive or negative integers which range from 0 to, or from 0 to.

                                                                                       (6-1)

An autocorrelation sequence A can thus be represented by a series of autocorrelation coefficients  obtained at every phase shift m. A raw signal sequence containing a periodic pattern can produce an autocorrelation sequence with its coefficient peaks where the phase shifts are equivalent to the periodicity of the raw signal sequence. Because the phase shifts can be either positive or negative from zero, the pattern of autocorrelation sequence [,,…, , ,…, , ] from the phase shift  to  can be symmetrical with zero phase shift located centrally (m=0). The autocorrelation sequence can either be “biased” or “unbiased”. The biased autocorrelation sequence  can be computed by Equation (6-2) that all the autocorrelation coefficients  are divided by the number the of raw signal sequence N. Equation (6-3) generates the unbiased autocorrelation  by dividing the autocorrelation coefficient  by the number. Figure 6.1 depicts the biased and unbiased autocorrelation sequences obtained from a raw acceleration signal measured from a young subject. Note that the number of autocorrelation coefficients equals  when the phase shift ranges from negative to positive.

                                                                                       (6-2)

                                                                             (6-3)

autocor_demo.jpg

Figure 6.1 The raw acceleration signal (above) and its biased (middle) and unbiased (below) autocorrelation sequences normalized to 1.0 at zero phase shifts

In Figure 6.1, both the biased and unbiased autocorrelation sequences differ very little at the center (zeroth phase shift) and the neighboring coefficients next to the center. However the signals differ obviously other than the central part. When the phase shift  increases, the number of the products  summed in its autocorrelation coefficient  is , and the value of  is smaller. Therefore, the values of the  will attenuate in the biased pattern because every autocorrelation coefficients are divided by a constant number N. The autocorrelation coefficients are only divided by  in an unbiased pattern and therefore they will not attenuate obviously until the pattern deteriorates at the both end edges. The unbiased method is preferred because the biased method generates noticeable attenuation of coefficient values next to the zero phase shift from a limited number of data. (Note that the MATLAB offers two autocorrelation functions: “xcorr” and “xcov”. The function “xcov” removes the average of the data sequence before calling “xcorr”. This is suggested by Moe-Nilssen et al. [2004] that this method rejects signal offset and would be mathematically suitable for algorithm in use.)

6.2.2 Interpreting gait cycle parameters from autocorrelation coefficient sequence

Figure 6.2 shows an example of unbiased autocorrelation pattern segment computed from the vertical acceleration sequence (measured from the waist, right ilium) during walking. Because the entire pattern is symmetrical with the zero phase shift, only the right half part (i.e., the phase shift from m=0 to m=) of entire pattern normalized to 1.0 at its zero phase shift is considered.

The coefficient at the zeroth phase shift always has maximal amplitude over the entire autocorrelation sequence because this zero phase shift point indicates the comparison of the original signal sequence to itself. By increasing the phase shift m, the first coefficient peak next to the zero phase shift point can be identified. The length n between the zeroth phase shift point and D1 implies that the original signal sequence has most apparent cyclic pattern with two signal points of n-sampled spaced. The phase shift n is the duration of the first dominant period which corresponds to one step. Similarly, the second peaks D2 indicates the second dominant period of the original signal sequence and it corresponds to a stride [Moe-Nilssen et al., 2004].

Figure 6.2 The example of an autocorrelation sequence computed from the vertical acceleration measured at waist during walking

(1)  Cadence estimation

Cadence is defined as the number of steps taken per minute. The duration of dominant period n can be used to derive estimated cadence. Assume M the number of steps per distance D walked, and V the walking speed (the distance D divided by the time spent). Therefore, the cadence . Let N be the number of signal samples per distance D  and f the sampling frequency such that  and . The cadence can be substituted as: . In other words, the number n is the single variable to estimate cadence as the sampling frequency is invariable.

Figure 6.3 shows the example of the autocorrelation patterns of gaits at normal walking speed (above) and festinating speed (below) from the same person. In this figure the normal gait speed produces the cadence of 103 (n=29) and the festinating gait cadence is 188 (n=16).

Figure 6.3 Autocorrelation patterns of gaits at normal walking speed (above) and festinating speed (below) of the same person

(2)  Gait regularity and symmetry

As mentioned previously, the first dominant period represents one step and the coefficient amplitude D1 can be regarded as the “step regularity”. Similarly, the coefficient amplitude D2 is the “stride regularity”. If the value is more close to 1.0 as its zeroth phase shift, the step or stride repeats more regularly from the signal sequence. Note that the right or left steps cannot be distinguished according to the autocorrelation pattern. However, the ratio  which represents the symmetry is defined and it is still valid [Moe-Nilssen et al., 2004]. Figure 6.4 shows the autocorrelation patterns obtained from normal gait (above) and pretended crippled gait (below) performed by a young adult. The step/stride regularity D1 and D2 of the normal gait are 0.7285 and 0.8342 which results in the gait symmetry of 0.8732. On the contrary, the crippled gait shows D1=0.1714 and D2=0.7550 such that its gait symmetry is 0.2270. The gait characteristics can be quantified and obviously the crippled gait shows lower gait symmetry even though both the cadences remain similar (103 and 100). Note that Figure 6.3 shows different cadences, however, both autocorrelation patterns show little difference in the step/stride regularity and symmetry (normal: D1=0.7285, D2=0.8342, symmetry=0.8732; festinating: D1=0.7130, D2=0.8871, symmetry=0.8037)) because both gaits were measured from the same person.

Figure 6.4 Autocorrelation patterns of normal gait (above) and pretended crippled gait (below)

6.3 Investigation of gait cycle parameters between healthy subjects and PD patients

In order to compare the gait cycle parameters between healthy subjects and PD patients, 10 subjects were recruited in a gait data collection in this study. Five subjects are healthy young males (26±3.1 yr) and without gait abnormalities, and the other 5 subjects (4 male and 1 female, 78±9.8 yr) are moderate PD patients diagnosed as Hoehn & Yahr (H&Y) stage II to III. For the PD participants, the test was conducted during their on-phase of the PD patients who had medication in the morning before the test. This data collection was approved by the Institutional Review Board at the Far-Eastern Memorial Hospital, Taipei. The participants were provided with necessary information about the test and they gave their informed consent before the test.

During the test, the participants wore the wearable motion detector and performed the Timed-Up and Go (TUG) test and 5-meter walk on a level ground. The accelerations of the test movements were recorded at a sampling rate of 50Hz by the wearable system and a camera beside the walkway recorded the synchronized video of the test movements for visual inspection of gaits. Each subject performed three TUG tests, three 5-meter walks at normal walking speed, and another three 5-meter walks at fast walking speed. The normal walking speed is self-regulated by the subjects, and is subject to the participants’ own normal and comfortable paces. For the fast walking speed, the subjects were asked to walk as they were in a hurry for something in their daily lives.

Figure 6.5 and Figure 6.6 show the examples of acceleration patterns measured from a PD subject (Figure 6.5) and a healthy subject (Figure 6.6) during walking at their normal self-regulated speed. By direct pattern inspection, the differences of gait parameters might not be obvious and interpretable. 

F6.5.jpg

Figure 6.5 The patterns of accelerations in the vertical and antero-posterior directions measured from a PD subject

fig6.7.jpg 

Figure 6.6 The patterns of accelerations in the vertical and antero-posterior directions measured from a healthy subject

Figure 6.7 and Figure 6.8 show the autocorrelation sequences of the acceleration patterns from Figure 6.5 and Figure 6.6, respectively. In both the Figure 6.7 and Figure 6.8, the first and second dominant periods on both the autocorrelation sequences of the vertical (VT) and antero-posterior (AP) acceleration patterns almost align with each other at the very exact time. Comparing the autocorrelation patterns in Figure 6.7 and Figure 6.8, the autocorrelation patterns of the PD subject contain more signal fluctuations between each dominant period. Moreover, and the dominant peaks on the both VT and AP patterns vary greatly. As Figure 6.8 showing the autocorrelation pattern from a young healthy subject, the patterns look more smooth and monotonic. Typically only the autocorrelation sequence of the VT acceleration components is used for deriving gait cycle parameters. The autocorrelation sequence of AP acceleration components can be additionally used to better identify the exact peaks of the dominant periods.

Figure 6.7 An example of the autocorrelation patterns of accelerations along vertical (VT, blue) and anterior- posterior (AP, red) directions of a PD subject

Figure 6.8 An example of the autocorrelation patterns of accelerations along vertical (VT, blue) and anterior- posterior (AP, red) directions of a healthy subject

Table 6.1 shows the statistical results of the gait-related parameters from the 10 test subjects. The average values with its standard deviation of each parameter are shown here. Note that among the 5 PD subjects, 2 of them were unable to perform the 5-meter walks above their normal walking speed. Therefore, for the subjects’ safety only 3 PD subjects participated in the data collection of the 5-meter walks at fast waking speed. The measured data of all the test subjects is listed in Table A6.1 to Table A6.3 in the Appendix of this chapter.

Table 6.1 Gait cycle parameters between PD patients

 

PD

Healthy

TUG-T time

23.9±7.9s

10.6±2.2s

5-meter walk

(normal speed)

Step regularity

0.39±0.16

0.63±0.13

Samples of a step length

30±4.70

30.5±1.85

Stride regularity

0.43±0.20

0.80±0.09

cadence

102.2±15.20

98.6±5.8

symmetry

0.97±0.3

0.79±0.17

5-meter walk

(fast speed)

Step regularity

0.37±0.17

0.76±0.08

Samples of a step length

28.2±3.7

26.40±1.4

Stride regularity

0.47±0.12

0.80±0.08

cadence

108.1±15.60

113.9±6.2

symmetry

0.77±0.25

0.94±0.12

6.4 Discussion

In this chapter, the autocorrelation procedure is used to estimate gait cycle parameters from trunk acceleration signals measured by the wearable motion detector. Temporal gait parameters, such as cadence, step regularity, stride regularity and step symmetry can be extracted from the sequence of autocorrelation coefficients of the accelerations. These parameters can be used to estimate an individual’s gait cycle characteristics during walking.

In order to investigate the difference of the pattern of autocorrelation coefficients and the gait cycle parameters between healthy and PD individuals, trunk accelerations during 5-meter walks at normal and fast walking speeds were measured from 5 healthy subjects and 5 patients with Parkinson’s disease. The clinical assessment method Timed Up-and Go Test was conducted first to briefly screen the mobility of the test subjects. The TUG time taken by the healthy group is 10.6±2.2s while a longer time 23.9±7.9s is measured in the PD group. This simple estimate shows a degenerative mobility for the PD patients.

Comparing the cadences computed from the autocorrelation procedure, the PD group has the cadence slightly higher than the healthy group. However, for the PD group the cadence at fast 5-meter walks is 108.1±15.6 step/min., which is approximately 5.77% increased from their normal cadences. The healthy group has the cadence of 113.9±6.2 steps/min. at fast 5-meter walks, which is approximately 16.94% increased from their normal cadences. This indicates a limited performance margin for the PD group due to their degenerative mobility. Also note that the larger deviations in the TUG-T and cadences are seen in the PD group due to their varied mobility level.

For the regularity and symmetry of gaits, the PD group has the step regularity of 0.39±0.16 and the stride regularity of 0.43±0.2 during walks of normal speed. The healthy group has higher step regularity (0.61±0.14) and stride regularity (0.79±0.09). Similar performance can be observed in the walking at fast speed. Therefore, the PD patients have less regular performance during repeating step and stride performance compared with the healthy group. Note that the symmetry during their normal walking speed in the PD group is higher than that in the healthy group while the symmetry during fast walking in the PD group is lower than that in the healthy group. This mixed results regarding gait symmetry need further investigations.

Several studies have reported the use of the vertical accelerations for autocorrelation procedure [Moe-Nilssen et al. 2004; Keenan et al., 2005; Yang et al., 2010]. In this chapter the accelerations in the vertical (VT), antero-posterior (AP) and medio-lateral (ML) directions were computed by the autocorrelation procedure to examine which axis is most sensitive to motions and produces identifiable pattern related to the gait cycle parameters. With visual inspection from the data of the 10 test subjects, the AP and ML components are considered to exhibit less sensitive and least descriptive autocorrelation pattern than the VT component. Therefore the ML acceleration is not used in the autocorrelation analysis. However, the autocorrelation sequences of AP and ML accelerations were observed to exhibit reduced first dominant period on some cases though the second dominant period was distinct and determinable.

From visual inspection, the autocorrelation sequences computed from healthy subjects in this study tend to exhibit more uniform and monotonic pattern as Figure 6.6 showing the patterns of VT and AP accelerations from a healthy subject. Referring to Figure 6.5, some signal fluctuations exist between the dominant periods of the patterns of VT components from a PD patient. Accordingly it implies less regular movements between each step, which can be considered the results from ill-controlled motor behaviors. This requires more PD data to justify this observation.

The purpose of this chapter is to develop a wearable system for real-time recognition of Parkinsonian gaits. The proposed system is able to recognize different temporal gait characteristics from people with varied mobility level. The current focus of interest here is whether those selected gait cycle parameters can be discriminative between the PD patients and normal healthy people. The PD group and the healthy group show different characteristics for the gait cycle parameters though some details still needs further investigation. The gait cycle parameters can be used and a signal processing algorithm can be developed for a wearable system for real-time recognition of abnormal gaits, like shuffling, or festinating gaits from PD patents. This development is expected to benefit and assists ambulation rehabilitation of PD patients and personal tele-care applications.

Appendix

Table A6.1 Timed Up and Go test performances from all the subjects

Subjects

Time (sec.)

Healthy group

Subject 1

1

9.3

2

10.0

3

9.6

Subject 2

1

13.8

2

13.8

3

13.1

Subject 3

1

13.2

2

12.5

3

12.6

Subject 4

1

8.4

2

8.2

3

8.5

Subject 5

 

1

8.4

2

9.3

3

8.6

PD group

Subject 1

1

17.4

2

19.0

3

20.0

Subject 2

1

42.4

2

32.1

3

33.6

Subject 3

1

17.3

2

15.3

3

13.1

Subject 4

1

22.6

2

20.9

3

23.1

Subject 5

1

28.6

22

28.0

33

25.2

Table A6.2 Gait cycle parameters during normal walking speed computed from the vertical acceleration of all the subjects

Subjects

D1

n

D2

Cadence

Symmetry

Healthy group

Subject 1

1

0.7184

29

0.8828

103.4

0.81

2

0.7235

29

0.9248

103.4

0.78

3

0.7178

29

0.7994

103.4

0.90

Subject 2

1

0.6773

30

0.7019

100.0

0.96

2

0.6022

30

0.7316

100.0

0.82

3

0.6148

30

0.8822

100.0

0.70

Subject 3

1

0.7504

32

0.7087

93.8

1.06

2

0.8409

32

0.8108

93.8

1.04

3

0.8165

34

0.825

88.2

0.99

Subject 4

1

0.4698

33

0.6865

90.9

0.68

2

0.5677

33

0.7822

90.9

0.73

3

0.574

31

0.8162

96.8

0.70

Subject 5

 

1

0.4062

28

0.6841

107.1

0.59

2

0.4402

29

0.9342

103.4

0.47

3

0.5482

29

0.903

103.4

0.61

PD group

Subject 1

1

0.2647

29

0.3633

103.4

0.73

2

0.3642

32

0.4241

93.8

0.86

3

0.2573

36

0.4398

83.3

0.59

Subject 2

1

0.2725

29

0.2758

103.4

0.99

2

0.3975

29

0.2948

103.4

1.35

3

0.3922

27

0.3926

111.1

1.00

Subject 3

1

0.5911

24

0.7769

125.0

0.76

2

0.5993

25

0.7265

120.0

0.82

3

0.6343

24

0.7962

125.0

0.80

Subject 4

1

0.5171

35

0.496

85.7

1.04

2

0.5027

36

0.4723

83.3

1.06

3

0.4049

39

0.3371

76.9

1.20

Subject 5

1

0.2327

27

0.3496

111.1

0.67

2

0.081

30

0.0917

100.0

0.88

3

0.3562

28

0.2025

107.1

1.76

Table A6.3 Gait cycle parameters during fast walking speed computed from the vertical acceleration of all the subjects

Subjects

D1

n

D2

Cadence

Symmetry

Healthy group

Subject 1

1

0.647

24

0.7712

125.0

0.84

2

0.6778

24

0.8637

125.0

0.78

3

0.7007

24

0.8736

125.0

0.80

Subject 2

1

0.7542

25

0.8203

120.0

0.92

2

0.8255

27

0.8051

111.1

1.03

3

0.795

25

0.7728

120.0

1.03

Subject 3

1

0.7436

28

0.6378

107.1

1.17

2

0.7865

29

0.7448

103.4

1.06

3

0.8303

28

0.7985

107.1

1.04

Subject 4

1

0.8441

26

0.904

115.4

0.93

2

0.7337

27

0.8265

111.1

0.89

3

0.8276

26

0.9323

115.4

0.89

Subject 5

 

1

0.5661

24

0.7875

125.0

0.72

2

0.6566

26

0.7533

115.4

0.87

3

0.6985

26

0.8713

115.4

0.80

PD group

Subject 1

1

0.1615

29

0.2674

103.4

0.60

2

0.2297

31

0.4317

96.8

0.53

3

0.1268

31

0.4291

96.8

0.30

Subject 2

n/a

Subject 3

1

0.5322

23

0.5101

130.4

1.04

2

0.4727

24

0.5685

125.0

0.83

3

0.3667

23

0.4178

130.4

0.88

Subject 4

1

0.5941

31

0.6969

96.8

0.85

2

0.4105

31

0.505

96.8

0.81

3

0.4624

31

0.4382

96.8

1.06

Subject 5

n/a

References

Berg K.O., Wood-Dauphinee S.L., Williams J.I., Gayton D., 1989. “Measuring balance in elderly: preliminary development of an instrument,” Physio-therapy Canada, Vol. 41, pp. 304-311.

de Melo Roiz R., Azevedo Cacho E.W., Pazinatto M.M., Guimarães Reis J., Cliquet A., Barasnevicius-Quagliato E.M.A., 2010. “Gait analysis comparing Parkinson’s disease with healthy elderly subjects,” Arq Neuropsiquiatr, Vol. 68, No. 1, pp. 81-86.

Goetz G. C., Poewe, W., Rascol O., Sampaio C., Stebbins G. T, Counsell C., Giladi N., Holloway R. G., Moore C.G., Wenning G.K., Yahr M.D., Seidl L., 2004. “Movement Disorder Society Task Force on the Hoehn and Yahr Staging Scale: Status and Recommendations,” Movement Disorders, Vol. 19, No. 9, pp. 1020-1028.  

Fahn S., Elton R. L., 1987. “Members of the UPDRS Development Committee: Unified Parkinson’s Disease Rating Scale. In: Fahn S., Marsden C. D., Calne D. B., Lieberman A,” Recent Development in Parkinson’s Disease. Florham Park, NJ: Macmillan Healthcare Information, pp. 153-163.

Femery V., Moretto P., Renaut H., Thévenon A., Lensel G., 2004. “Measurement of plantar pressure distribution in hemiplegic children: changes to adaptive gait patterns in accordance with deficiency,” Clinical Biomechanics, Vol. 17, pp. 406-413.

Hausdorff J. M., 2009. “Gait dynamics in Parkinson’s disease: Common and distinct behavior among stride length, gait variability, and fractal-like scaling,” Chaos, Vol. 19, pp. 026113-1-026113-14.

Hoehn M. M., Yahr M. D., 1967. “Parkinsonism: onset, progression and mortality,” Neurology, Vol. 17, pp. 427-442.

Jovanov E., Wang E., Verhagen L., Fredrickson M., Fratangelo R., 2009. “defog-a real time system for detection and unfreezeing of gait of Parkinson’s patients,” Proceedings of the 31st Annual International IEEE EMBS Conference, Minneapolis, MN, USA.

Keenan D.B., Wilhelm F.H., 2005. “Classification of locomotor activity by acceleration measurement: validation in Parkinson disease,” Biomedical Sciences and Instrumentation, Vol. 41, pp. 329-334.

Lowry K. A., Smiley-Oyen A. L., Carrel A. J., Kerr J. P., 2008. “Walking stability using harmonic ration in Parkinson’s disease,” Movement Disorder, Vol. 24, No. 2, pp. 261-267.

Mathie M., Coster A.C.F., Lovell N.H., Celler B.G., 2004. “Accelerometry: providing an integrated, practical method for long-term, ambulatory monitoring of human movement,” Physiological Measurement, Vol. 25, pp. R1-R20.

Menz H.B., Latt M.D., Tiedemann A., Kwan M.M.S., Lord S.R., 2004. “Reliability of the GAITRite walkway system for the quantification of temporal-spatial parameters of gait in young and older people, ” Gait & Posture, Vol. 20, pp. 20-25. 

Moe-Nilssen R., Helbostad J. L., 2004. “Estimation of gait cycle characteristics by trunk accelerometry,” Journal of Biomechanics, Vol. 37, pp. 121-126.

Moore S.T. MAcDougll H.G., Ondo W.G., 2008. “Ambulatory monitoring of freezing of gait in Parkinson’s disease,” Journal of Neuroscience Methods, Vol. 167, pp. 340-348.

Nelson-Wong E., Howarth S., Winter D.A., Callaghan J.P., 2009. “Application of autocorrelation and crosscorrelation analyses in human movement and rehabilitation research,” Journal of Orthopaedic & Sports Physical Therapy, Vol. 30, No. 4, pp. 287-295

Podsiadlo D., Richardson S., 1991. “The timed “Up & Go”: a test of basic functional mobility for frial elderly persons, ” J Am Geriatr Soc., Vol. 39, pp. 142-148.

Yang C.C., Hsu Y.L., 2010. “A review of accelerometry-based wearable motion detector for physical activity monitoring,” Sensors, Vol. 10, pp.7772-7788.

Yang M.J., Zheng H., Wang H., McClean S., Hall J., Harris N., 2010. “Assessing accelerometer based gait features to support gait analysis for people with complex regional pain syndrome,” Proceedings of the 3rd International Conference on Pervasive Technologies Related to Assistive Environments, NY, USA.