On Gait Recognition With Smartphone Accelerometer, Essay Example

Abstract

Face recognition, and the gait of each can be used for personal identification. However, the gait of every person is very imperative in distinguishing one individual from another. Furthermore, studies that were conducted on gait analysis relied mostly on laboratory experiments. Besides, technology advancement has shade more light on the gait analysis in an attempt to advance personal identification by the use of smartphones. In recent proposals, researchers have based their study on identification properties that are installed in most of the smartphones that are available in the market. However, this analysis must be narrow to accommodate the internal features of the smartphone. Therefore, this study outlines the key elements of gait recognition using a smartphone.

Introduction

Researchers in the field of medicine [1] have found out that each has a specific and unique walking style that differentiates them from other people. Therefore, personal identification through the use of gait patterns has raised a curiosity in the minds of scientists[3]. Current studies on the use of gait patterns in smartphones have advanced from the first researches. Moreover, sophisticated scientific instruments are nowadays being used in gait analysis and human walk [4].

The modern scientific facilities used in the smartphones are of low-level technology, although they have been found to work satisfactorily [5]. Personal identification test carried on them have confirmed that they are a better option and a cost friendly. Therefore, the gait recognition feature in a smartphone is an advantage that offers a relatively high level of security in terms of accessing the mobile applications in a way that is convenient to the user [6-12]. However, the user may overlook the gravity of the privacy of information that is stored in a smartphone. A pin or a password may not guarantee maximum protection of mobile phone since they are mostly required when switching off the phone [14].

Studies have shown that the use of a password has become a tiresome task primarily due to typing. Therefore, gait recognition can be used so that a smartphone can identify when the user is walking. Moreover, gait recognition is advantageous because there will be no instances of stolen vital data, which is usually the case with a password. However, this authentication technology will have to be combined with another privacy application in case the user is not moving.

Nevertheless, this study systematically outlines the necessary procedures and describes the methods proposed for gait recognition application installed in a smartphone. The scientific experiment has been carried out on Samsung Galaxy S5 to back up the study. The necessary steps outlined by the leading research approaches are highlighted in figure one and are further discussed in this study. Data collection is the first step, and it explains the position of the smartphone for clarity of the acceleration signal. The second step is the cycle detection in which receive patterns inside the data are detected. Therefore, the cycles are processed in terms of time and frequency domain in the analysis stage.

Data collection

Most of the solutions that are provided for carrying smartphones during the time of experiments were almost similar. The user can move a smartphone in various ways:

• Some of the people wear it on a belt [6,10]
• Others carry it inside a pouch [8,9]
• However, it can also be put in the pants pocket [12,13]
• The user can also tie it in their correspondence of the pants pocket [11]

Few researchers have proposed different approaches to securing the gait recognition device. Several experiments have suggested that the device can be achieved over the part of the body that is at the center of the mass when the user is standing [5]. The data is made for acceleration and is collected using a 3-axis accelerometer application. The sensor repetitively measures the speed along the 3-vertical axes. According to research, the X-axis represents the vertical component, the Y-axis corresponds to the forward-backward motion, and the Z-axis relates to the right acceleration. Therefore the required information is collected by placing the Samsung Galaxy S5 in the pocket at the right back of the user.

Preprocessing

All the three constituents of the accelerations are assessed as one in some approaches [12,13], but only a part of them may be considered. The vertical and the forward-backward one are the most necessary components and are mostly evaluated jointly [6]. However, some researches take into account the forward-backward [8,9] and vertical elements [10,11]. Most proposals assess the absolute value of the final vector resulting from the summation of the components, although it is made to avoid the influence of the smartphone features. Therefore, each free vector can be computed as follows [12,13]:

The interval of the data that is processed depends on the software used on the smartphone. Therefore, a linear interpolation of time is required to provide a uniform range [5,8-13]: and  are the accelerations at times  and , respectively. Also,  is the point from the interpolation of  and . The measure collected accounts for both movement and gravitational force of the earth. However, the gravitational force is usually omitted in some analysis. The gravity affects the alignment of the axes, especially when the mobile phone is kept vertically concerning the ground. Therefore, this needs correction which accomplished by subtracting the mean value ? [8]

The data collected may contain some disturbance component that is as a result of displacement of the smartphone caused by movements of the user. Some analysis evaluates the acceleration data after passing it through a series of filters [10].

Cycle detection

A cycle is represented by two systematic phases that can be identified using local minimums and maximums. It is measured using the length between two consecutive highest points of the maximum displacement. Some researchers have indicated that the range may not be apparent for the detection of cycles. Therefore, several studies have proposed that a predetermined average field is required for the precise detection of a sequence. However, sometimes, estimations can be used to determine the average cycle length. First, the center of the acceleration waveform is identified, and then a small subset is located, which is then compared to other subgroups of a similar length from the same waveform.

Salience vectors [8,9] are also used to evaluate the average length of a cycle. A normal salience vector contains data entry that corresponds to the acceleration samples. The entries that have values more significant than the least peak height but are equal to the least peak displacement to the cycle. The length of the period is computed at the mean value of the ranges between the candidate point. Therefore, the sequences are detected by enumerating the data points before a maximum amount regarding the least peak. The horizontal component of the acceleration is used as a reference point for discovering the starting and the ending point of the cycles. However, both the original and the smoothened data can be used to detect the sequences. Besides, real peaks can be identified in the first acceleration wavelength by observing data points around the candidate’s height.

Analysis in the time domain

However, this type of study deals with the sign spreading in a given period. When the investigation is used in gait recognition, it shows how every person faces effect concerning the data collected from the sample distribution in the accelerometer with the 3-axis.

Statistics of the cycle

After accelerating then samplings into d cycles, the data from the periods generates the explanation of the gait (12).

Moments

The following is a definition of the k-th flash of an unplanned separate adaptable x about adaptable c.

Where the outcome mass of function x is represented by pi. The usefulness of the instant unplanned separate acceleration of the gait is achieved through its ability to show the distribution of the data. The cycles that are applied show different data. However, the indication of the gait variable can be explained by the first and the second moment where the next moment shows the variance. Moreover, the square root that is obtained from the variation is the standard deviation (5).

Nevertheless, the other discriminant parameters of the acceleration waveform are given by the first and the second instants. Skewness is the other moment that shows the data measurement by the use of mean. Besides, kurtosis is the fourth-moment showing data. Also, it shows how the data is peaked to display a reasonable spread.

Kurtosis and Skewness can be represented graphically by use of the histogram. A histogram is drawn after the disjoint of the speeding up waveform. The number of values that are found in the bin gives the height of the container itself. The histogram displays the relative frequencies since the length shown by the benefits normalizes it.

Root Mean Square

The sample that is collected form values that are positive and others that are negative in the acceleration. RMS enables the obtaining of the portion of the degree.

However, the stability of the gait is achieved by the RMS. When the RMS is high, then the degree of confidence is low (5)

Correlation

The assumption that the shape of acceleration is unique to different people enables the gait correlation to cross-sect. Two values x and y are compared by cross-correlation through shifting them a long period of domain.

Autocorrelation is the process of comparing a signal correlation against itself. When the signal is at its t moment, it can be compared to its value at t+t

Extrapolating the data may be achieved by varying r:

• The speed the signal takes to change via period allocated. When there is the slow movement of the signal there is similarity in the values of xt and xt+r and there is also positive autocorrelation.
• The periodic patterns are available and the noise hides them.
• The proof of identity of a significant frequency of the sign that has a frequency that is cool instead of a component that is noise.

The varying time lag of r leads to the computing of the autocorrelation coefficient. The estimation of the gait balance is achieved from the provision of the maximum value of the gait. For the fast estimation of the coefficient, each value is calculated as:

Where the value of n represents the number of units accelerating

The computation of the covariance of the values of r helps in the construction of the covariance matrix. The covariance of x forms the variance of  .

Dynamic time warping

DTW can be described as an algorithm that is used in the comparison of two different signals X=(,,…,) and y=(,,…,) [15]. For the evaluation of similarity between elements of a signal, there must be a definition of the distance measure d(, ). DTW aims at calculating the path of warping by use of the least cumulative distance. (1,1) is the start point for the path and (N,M) is the endpoint. The following rules are applied when corresponding to the warping path and optimal match.

• The element that is at the first position is equated with the first one of y and that also happens to the last elements that belong to the signals.
• For compensation of the non- linear variations, there must be wrapping of the signal in the time dimension.
• There should be no omission of the element in the match.

Analysis in the frequency domain

The analysis reveals how much is diverted from the signal to the bands by the use of the different ranges of the frequency. Transform is the mathematical way of transferring the signal to within the domain of time. For a good study of the supply of the components, gait recognition is applied in the acceleration wavelength. The information that is obtained from the supply is extracted from each input of the study.

Discrete Fourier transform (DFT)

Discrete Fourier transformation helps in the conversion of the acceleration of the samples by converting them to the frequency domain. However, the samples must be spacious to give room for enough range of the collection in the intervals.

The first output values of the DFT contain large information that shows the gait characteristics. The fast Fourier transformation enables achievement of the speed by limiting the number of coefficients [6,7,8].

Discrete cosine transform (DCT)

DCT is useful in the conversion of the number of data into functions that are cosine. Different variants are applicable when obtaining the normal coefficients of DCT.

Discussion

The recognition of the gait through the state of the art proposal helps in the giving of guidelines that enable data collection. However, the first thing is selecting and polishing the data. after the sampling of the acceleration in perpendicular axes, the data that is collected is enabled dropping some of the values. Also, the measures should be taken into account when doing data normalization. Besides, the next step is the recognition of the periodicity of the information. A cycle is a pattern that corresponds to two steps. The cycles are then recognized through comparing them with templates where authentication is done. The use of a mathematical transformation enables conversion of the acceleration data. The frequency components are one of the characteristics of the gait.

While the sample is in the gait recognition, it is accepted only if the template is higher than that of the threshold. False acceptance rate and false rejection rates are the two types of errors that can occur. The increase in the threshold leads to the increase in false acceptance rate and hence leading to the growth of the false rejection rate. There is an impossible comparison of the two thresholds because of the difference in scenarios.

Conclusions

The analysis that is found in this paper explains how the gait can be recognized through acquiring data by use of a smartphone which has a 3-axis accelerometer. Different stages are identified in the process and include acquisition, preprocessing, cycle detection and analysis. Besides, there is a presentation of the technique that is used in the analysis. The approach that is used gives results that are viable.

Work Cited

[1] T. Öberg, A. Karsznia, and K. Öberg. “Basic gait parameters: reference data for normal subjects, 10-79 years of age”, Journal of rehabilitation research and development, vol. 30, issue 2, 1993, pp 210–223

[2] S. V. Stevenage, M. S. Nixon, and K. Vince, “Visual Analysis of Gait as a Cue to Identity”, Applied Cognitive Psychology, vol. 13, pp. 513–526, Dec. 1999, doi: 10.1002/(SICI)1099-0720(199912)13:63.0.CO;2-8

[3] D. Gafurov, “A survey of biometric gait recognition: approaches, security and challenges”, in Annual Norwegian Computer Science Conference, Nov. 2007

[4] J.-A. Lee, S.-H. Cho, J.-W. Lee, K.-H. Lee, and H.-K. Yang, “Wearable accelerometer system for measuring the temporal parameters of gait”, in 29th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBS), pp.483–486, Aug. 2007, doi: 10.1109/IEMBS.2007.4352328

[5] S. Nishiguchi, M. Yamada, K. Nagai, S. Mori, Y. Kajiwara, T. Sonoda, K. Yoshimura, H. Yoshitomi, H. Ito, K. Okamoto, T. Ito, S. Muto, T. Ishihara, and T. Aoyama, “Reliability and validity of gait analysis by android-based smartphone”, Telemedicine and e-Health, vol. 18, issue 4, May 2012, pp 292–296, doi:10.1089/tmj.2011.0132

[6] J. Mäntyjärvi, M. Lindholm, E. Vildjiounaite, S.–M. Mäkelä, and H. Ailisto, “Identifying users of portable devices from gait pattern with accelerometers”, in Proceedings of the IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP), vol.2, pp.ii/973–ii/976, Mar. 2005, doi: 10.1109/ICASSP.2005.1415569

[7] M. Tamviruzzaman, S. I. Ahamed, C. S. Hasan, and C. O’brien, “ePet: when cellular phone learns to recognize its owner”, in Proceedings of the 2nd ACM workshop on Assurable and usable security configuration (SafeConfig), pp. 13-18, Nov. 2009, doi: 10.1145/1655062.1655066

[8] M. Muaaz, and R. Mayrhofer, “An analysis of different approaches to gait recognition using cell phone based accelerometers”, in Proceedings of International Conference on Advances in Mobile Computing & Multimedia (MoMM), ACM, pp. 293–300, Dec. 2013, doi: 10.1145/2536853.2536895

[9] C. Nickel, M. O. Derawi, P. Bours, and C. Busch, “Scenario test of accelerometer-based biometric gait recognition”, in 3rd International Workshop on Security and Communication Networks (IWSCN), IEEE, pp.15–21, May 2011, doi: 10.1109/IWSCN.2011.6827712

[10] M.O. Derawi, C. Nickel, P. Bours, and C. Busch, “Unobtrusive user- authentication on mobile phones using biometric gait recognition”, in 6th International Conference on Intelligent Information Hiding and Multimedia Signal Processing (IIH-MSP), pp.306–311, Oct. 2010, doi: 10.1109/IIHMSP.2010.83

[11] H. M. Thang, V. Q. Viet, N. D. Thuc, and D. Choi, “Gait identification using accelerometer on mobile phone,” in International Conference on Control, Automation and Information Sciences (ICCAIS), Nov. 2012, pp.344–348, doi: 10.1109/ICCAIS.2012.6466615

[12] H. M. Thang, N. D. Thuc, C. Luong, S. Do, and D. Choi, “Adaptive cross-device gait recognition using a mobile accelerometer”, Journal of Information Processing Systems, vol. 9, issue 2, Jun. 2013, pp 333–348, doi: 10.3745/JIPS.2013.9.2.333

[13] F. Juefei-Xu, C. Bhagavatula, A. Jaech, U. Prasad, and M. Savvides, “Gait-ID on the move: Pace independent human identification using cell phone accelerometer dynamics,” in IEEE 5th International Conference on Biometrics: Theory, Applications and Systems (BTAS), Sept. 2012, pp. 8–15, doi: 10.1109/BTAS.2012.6374552

[14] F. Breitinger and C. Nickel, “User survey on phone security and usage”, in Proceedings of the Special Interest Group on Biometrics and Electronic Signatures (BIOSIG), pp. 139–144, Sep. 2010

[15] M. Müller, “Information retrieval for music and motion”, Springer- Verlag New York, USA, 2007, ch. 4 Dynamic Time Warping, pp. 69–84