A Machine Learning Strategy for Locomotion Classification and Parameter Estimation Using Fusion of Wearable Sensors

Abstract

The accurate classification of ambulation modes and estimation of walking parameters is a challenging problem that is key to many applications. Knowledge of the user's state can enable rehabilitative devices to adapt to changing conditions, while in a clinical setting it can provide physicians with more detailed patient activity information. This study describes the development and optimization process of a combined locomotion mode classifier and environmental parameter estimator using machine learning and wearable sensors. A detailed analysis of the best sensor types and placements for each problem is also presented to provide device designers with information on which sensors to prioritize for their application. For this study, 15 able-bodied subjects were unilaterally instrumented with inertial measurement unit, goniometer, and electromyography sensors and data were collected for extensive ranges of levelground, ramp, and stair walking conditions. The proposed system classifies steady state ambulation modes with 99% accuracy and ambulation mode transitions with 96% accuracy, along with estimating ramp incline within 1.25 degrees, stair height within 1.29 centimeters, and walking speed within 0.04 meters per second. Mechanical sensors (inertial measurement units, goniometers) are found to be most important for classification, while goniometers dominate ramp incline and stair height estimation, and speed estimation is performed largely with a single inertial measurement unit. The feature tables and Matlab code to replicate the study are published as supplemental materials.

Can humans perceive the metabolic benefit provided by augmentative exoskeletons?

“…Thus, we inferred the stride completion percentage using heel-strike events. We detected these events by thresholding the onboard accelerometers (MPU-9250, Invensense, San Jose, CA) [ 46 , 47 ].…”

IMU-Based Deep Neural Networks: Prediction of Locomotor and Transition Intentions of an Osseointegrated Transfemoral Amputee

“…2) Locomotor and Transition Intention Joint Prediction: This study focuses on the joint intention prediction of eight locomotion modes and twenty-four transitions of one osseointegrated transfemoral amputee. Previous research has been devoted to the disjoint prediction of five locomotion modes and nine transitions of healthy subjects with (several) feature engineering methods [5], to the prediction of six locomotion modes and twelve transitions of healthy subjects with a feature engineering method (LDA) on IMU data combined with pressure insoles [7], to the disjoint prediction of three locomotion modes and four transitions of healthy subjects with a features engineering method on one IMU [10], to the disjoint prediction of four locomotion modes and four transitions of healthy subjects with different features engineering methods on several IMUs and other sensors [12], and to the recognition of five locomotion modes and eight transitions of healthy subjects and transtibial amputees with feature learning methods (CNNs) [17].…”

“…3) Accuracy: This study shows that a RNN, realized with four layers of gated recurrent unit networks, achieves (with a 5-fold cross-validation) a mean F1 score of 93.06% (standard deviation of 1.21) using time and time-frequency domain features engineered from two IMUs, in the best preforming experimental scenarios (eight locomotion modes and twentyfour transitions). Previous research on the prediction of both locomotor and transition intentions has reported 97.65% on healthy subjects (five locomotion modes and nine transitions) with engineered features in the time-domain from seven IMUs [5], 99.71% on healthy subjects (six locomotion modes and twelve transitions) with engineered features in the time-domain from IMUs and pressure insoles [7], 98.7% on healthy subjects (three locomotion and four transitions) with engineered features from raw-data of one IMU [10], 99% on healthy subjects (four locomotion and four transitions) with engineered features from raw-data of several IMUs and several other sensors [12], 94.15% on healthy subjects (five locomotion modes and eight transitions) with features learned from raw-data of three IMUs [17], and 89.23% on transtibial amputees (five locomotion modes and eight transitions) with features learned from raw-data of three IMUs [17]. locomotor intention prediction of ten healthy subjects [16].…”

“…A binary tree is used on raw data from one IMU to predict locomotion modes and transitions of healthy subjects in [10], while a gradient tree boosting method on several time-domain IMU data in combination with encoders and load cells is proposed in [11] for the locomotion prediction of transfemoral amputees. In [12], different methods are compared for the recognition of locomotion modes and transitions of healthy subjects by using IMUs is combination with several sensors.…”

Locomotion Mode Transition Prediction Based on Gait-Event Identification Using Wearable Sensors and Multilayer Perceptrons

“…Locomotion modes are typically labeled between the FC and TO, which, in turn, are often identified through motion capture systems [23], load cells [24] and IMUs [25,26]. However, motion capture systems generally require a specialized lab, which is not always feasible.…”