Leveraging walking inertial pattern for terrain classification

2007

In this work, 33 dimensional time-frequency domain features were developed and evaluated to detect five different human walking patterns from data acquired using a triaxial accelerometer attached at the waist above the iliac spine. 52 subjects were asked to walk on a flat surface along a corridor, walk up and down a flight of a stairway and walk up and down a constant gradient slope, in an unsupervised manner. Time-frequency domain features of acceleration data in anterior-posterior (AP), medio-lateral (ML) and vertical (VT) direction were developed. The acceleration signal in each direction was decomposed to six detailed signals at different wavelet scales by using the wavelet packet transform. The rms values and standard deviations of the decomposed signals at scales 5 to 2 corresponding to the 0.78-18.75 Hz frequency band were calculated. The energies in the 0.39-18.75 Hz frequency band of acceleration signal in AP, ML and VT directions were also computed. The back-end of the system was a multi-layer perceptron (MLP) Neural Networks (NNs) classifier. Overall classification accuracies of 88.54% and 92.05% were achieved by using a round robin (RR) and random frame selecting (RFS) train-test method respectively for the five walking patterns.

IFAC-PapersOnLine, 2020

In this paper, we present a method to obtain explicit, expressive and interpretable gait feature signals from an inertial sensor, mounted on any segment of the lower limbs. The proposed method is invariant to the mounting orientation of the sensor, works without magnetometer information, requires no prior knowledge and can be used in real-time scenarios. Moreover, the constructed signals are robust for a wide variety of changing walking speeds and directions. We investigate the informational content of our three feature signals lying in the human sagittal plane with respect to the gait phase segmentation problem and compare them to other commonly used signals, such as the sagittal angular velocity and the norms of accelerations and angular velocities. To this end, we make use of the filter-based maximum relevance minimum redundancy algorithm, which is a classifier-independent feature selection method. For validating our approach, we consider gait data of twelve healthy subjects walking straight and in curves at self-chosen speeds with inertial sensors attached to either the thigh, shank or foot. Additionally, pressure measuring insoles are used to obtain ground truth toe-off and heel-strike gait events for reference. With those events as the gait phase transitions, the event detection is cast into a classification problem. To support the theoretical findings of the feature selection and ranking, we finally evaluate different choices of feature sets with a simple linear support vector machine classifier in an online fashion and obtain superior segmentation results with our feature signals.

2011 IEEE International Conference on Signal and Image Processing Applications (ICSIPA), 2011

This paper presents an investigation on how different walking surfaces affect gait identification based on accelerometer data. Accelerometer data of 5 subjects walking on 4 different solid surfaces was acquired on 3 different days by a cell phone placed on the subject's hip. Data analysis detects gait cycles by a method based on wavelet transform. For all gait cycles, features were calculated by using higher-order statistics. Similarity estimation for discerning the different subjects and surfaces was introduced by using principal component analysis (PCA). It proved that the different solid surfaces do not influence the efficiency of the identification of subjects based on their gait.

2020

Machine learning and data analytics are becoming pervasive also in the analysis of human behaviour as more and more miniaturized sensors can observe and quantify human activities. The so-called wearables are particularly tiny transducers which give the possibility to understand an individual's behaviour possibly enabling innovative services such as gait analysis based identification, foot pressure analysis, fall prevention and automatic recognition of dangerous situations. In this work we investigate an innovative experimental setting where accelerations are captured, during walking, by an Inertial Measurement Unit embedded in the shoe's sole. The goal is to identify who is wearing the shoe by simply analyzing his gait. User classification is then performed comparing different machine learning methods, relying either on the k-Neareast-Neighbour or on the Linear Discriminant Analysis algorithm. An extensive experimental campaign was carried out on five young adults and a comp...

IEEE Transactions on Biomedical Engineering, 2017

For healthcare and clinical use, ambulatory gait monitoring systems using inertial sensors have been developed to estimate the user gait parameters, such as walking speed, stride time, and stride length. However, to adapt the systems effectively to daily-life activities, they need to be able to classify the gait activities of daily-life to obtain the parameters for each activity. In this study, we propose a simple classification algorithm based on a single inertial sensor for ease of use, which classifies three major gait activities: leveled walk, ramp walk, and stair walk. Method: The classification can be performed with gait parameter estimation simultaneously. The developed system that includes classification and parameter estimation algorithms was evaluated with eight healthy subjects within a gait lab and on an outdoor daily-life walking course. Results: The results showed that the estimated gait parameters were comparable to existing studies (range of walking speed root mean square error (RMSE): 0.059-0.129 m/s), and the classification accuracy was sufficiently high for all three gait activities: 98.5 % for the indoor gait lab experiment and 95.5 % for the outdoor complex daily-life walking course experiment. Conclusion: The proposed system is simple and effective for daily-life gait analysis, including gait activity classification and gait parameter estimation. for each activity.

IEEE Transactions on Biomedical Engineering, 2005

Ageing influences gait patterns causing constant threats to control of locomotor balance. Automated recognition of gait changes has many advantages including, early identification of at-risk gait and monitoring the progress of treatment outcomes. In this paper, we apply an artificial intelligence technique [support vector machines (SVM)] for the automatic recognition of young-old gait types from their respective gait-patterns. Minimum foot clearance (MFC) data of 30 young and 28 elderly participants were analyzed using a PEAK-2D motion analysis system during a 20-min continuous walk on a treadmill at self-selected walking speed. Gait features extracted from individual MFC histogram-plot and Poincaré-plot images were used to train the SVM. Cross-validation test results indicate that the generalization performance of the SVM was on average 83.3% ( 2 9) to recognize young and elderly gait patterns, compared to a neural network's accuracy of 75 0 5 0%. A "hill-climbing" feature selection algorithm demonstrated that a small subset (3-5) of gait features extracted from MFC plots could differentiate the gait patterns with 90% accuracy. Performance of the gait classifier was evaluated using areas under the receiver operating characteristic plots. Improved performance of the classifier was evident when trained with reduced number of selected good features and with radial basis function kernel. These results suggest that SVMs can function as an efficient gait classifier for recognition of young and elderly gait patterns, and has the potential for wider applications in gait identification for falls-risk minimization in the elderly.

Journal of Medical Engineering & Technology, 2019

With the rising popularity of activity tracking, there is a desire to not only count the number of steps a person takes, but also identify the type of step (e.g., walking or running) they are taking. For rehabilitation and athletic training, this difference is important to the prescribed regiment. Fourteen healthy adults walked, jogged and ran on a treadmill at three different constant speeds (1.21, 2.01, 2.68 m/s) for 90 s. An inertial measurement unit (IMU) with accelerometer and gyroscope was affixed to their left ankle. Collected acceleration and angular velocity data were partitioned into individual time-normalised strides. These data were used as features in the artificial neural network (ANN) that classified the type of stride. Several ANN models were tested: using only acceleration, only angular velocity and both. Using primarily acceleration data in the trained ANN yielded the best results (>94% correct stride-type identification) after cross-validation. The ANN models were able to accurately classify the gait type of each stride using a single wearable IMU. The accuracy of the method should improve further as more data is added to the ANN training.

Robotica

Human gait data can be collected using inertial measurement units (IMUs). An IMU is an electronic device that uses an accelerometer and gyroscope to capture three-axial linear acceleration and three-axial angular velocity. The data so collected are time series in nature. The major challenge associated with these data is the segmentation of signal samples into stride-specific information, that is, individual gait cycles. One empirical approach for stride segmentation is based on timestamps. However, timestamping is a manual technique, and it requires a timing device and a fixed laboratory set-up which usually restricts its applicability outside of the laboratory. In this study, we have proposed an automatic technique for stride segmentation of accelerometry data for three different walking activities. The autocorrelation function (ACF) is utilized for the identification of stride boundaries. Identification and extraction of stride-specific data are done by devising a concept of tunin...

Computational and Mathematical Methods in Medicine, 2013

Two approaches to the classification of different locomotor activities performed at various speeds are here presented and evaluated: a maximum a posteriori (MAP) Bayes' classification scheme and a Support Vector Machine (SVM) are applied on a 2D projection of 16 features extracted from accelerometer data. The locomotor activities (level walking, stair climbing, and stair descending) were recorded by an inertial sensor placed on the shank (preferred leg), performed in a natural indoor-outdoor scenario by 10 healthy young adults (age 25-35 yrs.). From each segmented activity epoch, sixteen features were chosen in the frequency and time domain. Dimension reduction was then performed through 2D Sammon's mapping. An Artificial Neural Network (ANN) was trained to mimic Sammon's mapping on the whole dataset. In the Bayes' approach, the two features were then fed to a Bayes' classifier that incorporates an update rule, while, in the SVM scheme, the ANN was considered as the kernel function of the classifier. Bayes' approach performed slightly better than SVM on both the training set (91.4% versus 90.7%) and the testing set (84.2% versus 76.0%), favoring the proposed Bayes' scheme as more suitable than the proposed SVM in distinguishing among the different monitored activities.

2007 15th International Conference on Digital Signal Processing, 2007

The analysis of gait data has been a challenging problem and several new approaches have been proposed in recent years. This paper describes a novel front-end for classification of gait patterns using data obtained from a tri-axial accelerometer. The novel features consist of delta features, low and high frequency signal variations and energy variations in both frequency bands. The back-end of the system is a Gaussian mixture model based classifier. Using Bayesian adaptation, an overall classification accuracy of 96.1% was achieved for five walking patterns in six subjects.