Steady Locomotion Research Articles

Migration of confined micro-swimmers subject to anisotropic diffusion

Shear-induced migration of elongated micro-swimmers exhibiting anisotropic Brownian diffusion at a population scale is investigated analytically in this work. We analyse the steady motion of confined ellipsoidal micro-swimmers subject to coupled diffusion in a general setting within a continuum homogenisation framework, as an extension of existing studies on macro-transport processes, by allowing for the direct coupling of convection and diffusion in local and global spaces. The analytical solutions are validated successfully by comparison with numerical results from Monte Carlo simulations. Subsequently, we demonstrate from the probability perspective that symmetric actuation does not yield net vertical polarisation in a horizontal flow, unless non-spherical shapes, external fields or direct coupling effects are harnessed to generate steady locomotion. Coupled diffusivities modify remarkably the drift velocity and vertical migration of motile micro-swimmers exposed to fluid shear. The interplay between stochastic swimming and preferential alignment could explain the diverse concentration and orientation distributions, including rheological formations of depletion layers, centreline focusing and surface accumulation. Results of the analytical study shed light on unravelling peculiar self-propulsion strategies and dispersion dynamics in active-matter systems, with implications for various transport problems arising from the fluctuating shape, size and other external or inter-particle interactions of swimmers in confined environments.

Real-Time Continuous Locomotion Mode Recognition and Transition Prediction for Human With Lower Limb Exoskeleton.

Real-time continuous locomotion mode recognition and seamless timely transition detection is critical for the exoskeleton robot. This study aims to present a comprehensive and innovative framework for locomotion mode recognition and transition prediction, exclusively utilizing inertial measurement unit (IMU) signals from the exoskeleton. In this framework, a CNN-BiLSTM model was developed and trained to be the classifier and a novel majority filter was designed to reduce the transition misjudgment rate. Moreover, a comprehensive evaluation system encompassing eight dimensions for the classifier, incorporating evaluation metrics specifically for transition misjudgment, was proposed. We collected locomotion motion data from six subjects wearing a rigid exoskeleton robot using six IMU sensors on the exoskeleton. The proposed method achieves a high level of recognition accuracy, with an overall average of 99.58% for the five steady locomotion modes (level ground walking (LG), stair ascent/descent (SA/SD), and ramp ascent/descent (RA/RD)) across six subjects following the transition decision. All transitions are recognizable, and the majority can be predicted in advance, with an average prediction time of 353ms. Furthermore, the implementation of majority filter resulted in an average 87.04% reduction in the transition misjudgment rate among six subjects, thereby decreasing the average transition misjudgment rate to 4.82%. Finally, the model was tested on a Jetson Nano to verify its real-time performance. The results presented above were obtained under the condition where either leg could function as the first transition leg and revealed that the developed system was capable of achieving precise locomotion mode recognition and timely transition prediction, with high real-time performance.

Understanding muscle function during perturbed in vivo locomotion using a muscle avatar approach.

The work loop technique has provided key insights into in vivo muscle work and power during steady locomotion. However, for many animals and muscles, ex vivo experiments are not feasible. In addition, purely sinusoidal strain trajectories lack variations in strain rate that result from variable loading during locomotion. Therefore, it is useful to develop an 'avatar' approach in which in vivo strain and activation patterns from one muscle are replicated in ex vivo experiments on a readily available muscle from an established animal model. In the present study, we used mouse extensor digitorum longus (EDL) muscles in ex vivo experiments to investigate in vivo mechanics of the guinea fowl lateral gastrocnemius (LG) muscle during unsteady running on a treadmill with obstacle perturbations. In vivo strain trajectories from strides down from obstacle to treadmill, up from treadmill to obstacle, strides with no obstacle and sinusoidal strain trajectories at the same amplitude and frequency were used as inputs in work loop experiments. As expected, EDL forces produced with in vivo strain trajectories were more similar to in vivo LG forces (R2=0.58-0.94) than were forces produced with the sinusoidal trajectory (average R2=0.045). Given the same stimulation, in vivo strain trajectories produced work loops that showed a shift in function from more positive work during strides up from treadmill to obstacle to less positive work in strides down from obstacle to treadmill. Stimulation, strain trajectory and their interaction had significant effects on all work loop variables, with the interaction having the largest effect on peak force and work per cycle. These results support the theory that muscle is an active material whose viscoelastic properties are tuned by activation, and which produces forces in response to deformations of length associated with time-varying loads.

Dynamic visual acuity during asymmetric walking

Necessary for effective ambulation, head stability affords optimal conditions for the perception of visual information during dynamic tasks. This maintenance of head-in-space equilibrium is achieved, in part, by the attenuation of the high frequency impact shock resulting from ground contact. While a great deal of experimentation has been done on the matter during steady state locomotion, little is known about how locomotor asymmetry might affect head stability or dynamic visual acuity. In this study, fifteen participants walked on a split-belt treadmill while verbally reporting the orientation of a randomized Landolt-C optotype that was projected at heel strike. Participants were exposed to baseline, adaptation, and washout conditions, as characterized by belt speed ratios of 1:1, 1:3, and 1:1, respectively. Step length asymmetry, shock attenuation, high and low frequency head signal power, and dynamic visual acuity were averaged across the first and last fifty strides of each condition. Across the first fifty strides, step length asymmetry was significantly greater during adaptation than during baseline (p<0.001; d=2.442), and shock attenuation was significantly lower during adaptation than during baseline (p=0.041; d=-0.679). High frequency head signal power was significantly greater during adaptation than during baseline (p<0.001; d=-1.227), indicating a reduction in head stability. While dynamic visual acuity was not significantly lower during adaptation than during baseline (p=0.052), a moderate effect size suggests a decrease in the measure between the two conditions (d=0.653). Across the last fifty strides, many of the decrements observed between the baseline and adaptation conditions were greatly reduced. The results of this study indicate that the locomotor asymmetry imposed by the split-belt treadmill during early adaptation might lead to moderate decrements in shock attenuation, head stability, and dynamic visual acuity. Moreover, the relative reduction in magnitude of these decrements across the last fifty strides underscores the adaptive nature of the locomotor and visuomotor systems.

Steady locomotion in torso-actuated rimless wheel robots

Steady locomotion in torso-actuated rimless wheel robots

Compact-design, Modeling and Simulation of the Internal Permanent Magnet of Magnetically Guided Capsule Endoscope

Miniaturization is a challenge for the magnetically guided capsule endoscope (MGCE), especially considering the multifunction of MGCE. This paper proposes a compact-design of internal permanent magnet (IPM), which contributes to the miniaturization of MGCE. The IPM is a complex of actuator and shell of the MGCE, which frees the occupation of conventional IPM. The IPM is for the first time designed with a streamline shape for good kinematics. A magnetic propulsion model of the proposed IPM is established and a magnetic propulsion simulation based on finite element (FE) and boundary element (BE) analysis for guaranteeing steady locomotion and navigation is conducted. The simulation is validated by comparison of the analytical model. At last, the locomotion modes and navigation of IPM are simulated based on the justified simulation. In conclusion, the compact-design of IPM can guarantee the locomotion and navigation of MGCE, and increase the available inner space of MGCE.

Terrain Adaptability and Optimum Contact Stiffness of Vibro-bot with Arrayed Soft Legs.

The terrain adaptability of the state-of-the-art robot is far behind natural animals, partly because of limited sensing, intelligence, controlling, and actuating ability. One possible solution is to explore the flexible locomotion structure and locomotion mode with good adaptability and fault tolerance. Based on this idea, we presented a type of vibro-bot with arrayed soft legs (VBASL) with excellent terrain adaptability by utilizing the rapid vibration of the soft belt array. With the resistance to local terrain blocking and combing the vibrational actuation, the VBASL has an advantage of multi-leg collaboration, so that very simple structure can achieve good terrain adaptability, such as steady locomotion on complex terrains like steep slope, ladders, steps, discrete pillars, and soft sands. Besides, the effects of soft leg geometry, stiffness, and ground topography on terrain adaptability and locomotion speed were also studied, indicating the similar contact stiffness to maximize the locomotion speed on different grounds. Then, a theoretical model was developed to describe the experiments well, which can guide the design of optimum contact stiffness of VBASL to achieve fast locomotion speed and good load capacity. By further modifying the robot structure, more practical functions such as turning, climbing, and anti-impacting were easily realized. The resistance to local terrain blocking and optimum contact stiffness are two important factors to improve the performance of VBASL, which may address the terrain adaptability challenge of robots working in practical unstructured environments.

Overall dynamic body acceleration measures activity differently on large versus small aquatic animals

Abstract Acceleration‐based proxies for activity and energy expenditure are widely used in bio‐logging studies of animal movement and locomotion to explore biomechanical strategies, energetic costs of behaviour, habitat use and the impact of anthropogenic disturbance. The foremost such proxy is overall dynamic body acceleration (ODBA) along with variants VeDBA and PDBA. This technique, which involves summing the magnitude of high‐pass‐filtered acceleration signals (the so‐called dynamic acceleration) over a reference interval, has been applied to animals as diverse as shags, lobsters, humans and whales. The relationship between ODBA and energy use has been tested empirically on animals small enough to house in laboratory facilities and arguments have been offered for why the method should be generally applicable; however, validations on larger animals are scant. Here, we examine how body size impacts ODBA and its variants under steady locomotion in large aquatic animals, using cetaceans as model species. To do this, we first develop a simplified mathematical model for the acceleration signals that would be measured by a tag on a swimming animal. We then test this model with empirical data gathered using bio‐logging tags on whale species spanning 1.3 m harbour porpoises to 12‐m sperm whales. We show that the motions measured by ODBA can be fundamentally different in small versus large aquatic animals. Whereas dynamic acceleration in small animals is predominantly due to specific acceleration (i.e. actual accelerative motions generated by muscle action), in larger aquatic animals, body rotations (i.e. cyclical angular displacements associated with swimming) can dominate the measured acceleration. As body rotations do not necessarily increase in magnitude as swimming speed increases, ODBA may underestimate the relative cost of behaviours or responses to disturbance in large aquatic animals. This does not lessen the value of ODBA for small animals, but it raises a caution against uncritical use on larger animals. For large aquatic animals, activity proxies that specifically remove body rotations using gyroscopes or magnetometers may provide more consistent estimates of energy use although these methods are yet to be validated.

Investigating the role of low level reinforcement reflex loops in insect locomotion

Insects are highly capable walkers, but many questions remain regarding how the insect nervous system controls locomotion. One particular question is how information is communicated between the ‘lower level’ ventral nerve cord (VNC) and the ‘higher level’ head ganglia to facilitate control. In this work, we seek to explore this question by investigating how systems traditionally described as ‘positive feedback’ may initiate and maintain stepping in the VNC with limited information exchanged between lower and higher level centers. We focus on the ‘reflex reversal’ of the stick insect femur-tibia joint between a resistance reflex (RR) and an active reaction in response to joint flexion, as well as the activation of populations of descending dorsal median unpaired (desDUM) neurons from limb strain as our primary reflex loops. We present the development of a neuromechanical model of the stick insect (Carausius morosus) femur-tibia (FTi) and coxa-trochanter joint control networks ‘in-the-loop’ with a physical robotic limb. The control network generates motor commands for the robotic limb, whose motion and forces generate sensory feedback for the network. We based our network architecture on the anatomy of the non-spiking interneuron joint control network that controls the FTi joint, extrapolated network connectivity based on known muscle responses, and previously developed mechanisms to produce ‘sideways stepping’. Previous studies hypothesized that RR is enacted by selective inhibition of sensory afferents from the femoral chordotonal organ, but no study has tested this hypothesis with a model of an intact limb. We found that inhibiting the network’s flexion position and velocity afferents generated a reflex reversal in the robot limb’s FTi joint. We also explored the intact network’s ability to sustain steady locomotion on our test limb. Our results suggested that the reflex reversal and limb strain reinforcement mechanisms are both necessary but individually insufficient to produce and maintain rhythmic stepping in the limb, which can be initiated or halted by brief, transient descending signals. Removing portions of this feedback loop or creating a large enough disruption can halt stepping independent of the higher-level centers. We conclude by discussing why the nervous system might control motor output in this manner, as well as how to apply these findings to generalized nervous system understanding and improved robotic control.

The Role of Surface Electromyography in Data Fusion with Inertial Sensors to Enhance Locomotion Recognition and Prediction.

Locomotion recognition and prediction is essential for real-time human–machine interactive control. The integration of electromyography (EMG) with mechanical sensors could improve the performance of locomotion recognition. However, the potential of EMG in motion prediction is rarely discussed. This paper firstly investigated the effect of surface EMG on the prediction of locomotion while integrated with inertial data. We collected EMG signals of lower limb muscle groups and linear acceleration data of lower limb segments from ten healthy participants in seven locomotion activities. Classification models were built based on four machine learning methods—support vector machine (SVM), k-nearest neighbor (KNN), artificial neural network (ANN), and linear discriminant analysis (LDA)—where a major vote strategy and a content constraint rule were utilized for improving the online performance of the classification decision. We compared four classifiers and further investigated the effect of data fusion on the online locomotion classification. The results showed that the SVM model with a sliding window size of 80 ms achieved the best recognition performance. The fusion of EMG signals does not only improve the recognition accuracy of steady-state locomotion activity from 90% (using acceleration data only) to 98% (using data fusion) but also enables the prediction of the next steady locomotion (∼370 ms). The study demonstrates that the employment of EMG in locomotion recognition could enhance online prediction performance.

Repeated exposure to tripping like perturbations elicits more precise control and lower toe clearance of the swinging foot during steady walking.

Controlling minimum toe clearance (MTC) is considered an important factor in preventing tripping. In the current study, we investigated modifications of neuro-muscular control underlying toe clearance during steady locomotion induced by repeated exposure to tripping-like perturbations of the right swing foot. Fourteen healthy young adults (mean age 26.4±3.1years) participated in the study. The experimental protocol consisted of three identical trials, each involving three phases: steady walking (baseline), perturbation, and steady walking (post-perturbation). During the perturbation, participants experienced 30 tripping-like perturbations at unexpected timing delivered by a custom-made mechatronic perturbation device. The temporal parameters (cadence and stance phase%), mean, and standard deviation of MTC were computed across approximately 90 strides collected during both baseline and post-perturbation phases, for all trials. The effects of trial (three levels), phase (two levels: baseline and post-perturbation) and foot (two levels: right and left) on the outcome variables were analyzed using a three-way repeated measures analysis of variance. The results revealed that exposure to repeated trip-like perturbations modified MTC toward more precise control and lower toe clearance of the swinging foot, which appeared to reflect both the expectation of potential forthcoming perturbations and a quicker compensatory response in cases of a lack of balance. Moreover, locomotion control enabled subjects to maintain symmetric rhythmic features during post-perturbation steady walking. Finally, the effects of exposure to perturbation quickly disappeared among consecutive trials.

High-speed crawling-like locomotion robot using wobbling mass and reaction wheel

Aiming at generating steady and speedy locomotion on a slippery road, the authors have been investigating crawling-like locomotion robot using inner-wobbling effect. This paper discusses gaining transfer efficiency on an underactuated locomotion robot, which consists of an arc-shaped body with both a telescopic joint and a reaction wheel at the center of mass. First, we show the robot model, and then derive the equation of motion. A hierarchical control is also proposed to stabilize the dynamics. Second, we show the simulation results of a speedy locomotion on a slippery level surface. Finally, we discuss the stability of the reaction wheel motion.

Minimalist analogue robot discovers animal-like walking gaits

Robots based on simplified or abstracted biomechanical concepts can be a useful tool for investigating how and why animals move the way they do. In this paper we present an extremely simple quadruped robot, which is able to walk with no form of software or controller. Instead, individual leg movements are triggered directly by switches on each leg which detect leg loading and unloading. As the robot progresses, pitching and rolling movements of its body result in a gait emerging with a consistent leg movement order, despite variations in stride and stance time. This gait has similarities to the gaits used by walking primates and grazing livestock, and is close to the gait which was recently theorised to derive from animal body geometry. As well as presenting the design and construction of the robot, we present experimental measurements of the robot’s gait kinematics and ground reaction forces determined using high speed video and a pressure mat, and compare these to gait parameters of animals taken from literature. Our results support the theory that body geometry is a key determinant of animal gait at low speeds, and also demonstrate that steady state locomotion can be achieved with little to no active control.

Comprehensive mechanical power analysis in sprint running acceleration.

Sprint running is a common feature of many sport activities. The ability of an athlete to cover a distance in the shortest time relies on his/her power production. The aim of this study was to provide an exhaustive description of the mechanical determinants of power output in sprint running acceleration and to check whether a predictive equation for internal power designed for steady locomotion is applicable to sprint running acceleration. Eighteen subjects performed two 20m sprints in a gym. A 35-camera motion capture system recorded the 3D motion of the body segments and the body center of mass (BCoM) trajectory was computed. The mechanical power to accelerate and rise BCoM (external power, Pext ) and to accelerate the segments with respect to BCoM (internal power, Pint ) was calculated. In a 20m sprint, the power to accelerate the body forward accounts for 50% of total power; Pint accounts for 41% and the power to rise BCoM accounts for 9% of total power. All the components of total mechanical power increase linearly with mean sprint velocity. A published equation for Pint prediction in steady locomotion has been adapted (the compound factor q accounting for the limbs' inertia decreases as a function of the distance within the sprint, differently from steady locomotion) and is still able to predict experimental Pint in a 20m sprint with a bias of 0.70±0.93Wkg-1 . This equation can be used to include Pint also in other methods that estimate external horizontal power only.

Evaluation of time-frequency features as detectors of lack of balance due to tripping-like perturbations.

Unbalancing events during gait can end up in falls and, thus, injury. Detecting events that could bring to fall and consequently activating fall prevention systems before the impact may help to mitigate related injuries. However, there is uncertainty about signals and methods that could offer the best performance. In this paper we investigated a novel trip detection method based on time-frequency features to evaluate the performances of these features as trip detectors. Hip angles of eight healthy young subjects were recorded while performing unexpected tripping trials delivered during steady locomotion. Then the Short-Time Fourier Transform (STFT) of the hip angle was estimated. Median frequency, power, centroidal frequency as well as frequency dispersion were computed for each time sliced power spectrum. These features were used as input for a trip detection algorithm. We assessed detection time (Tdetect), specificity (Spec) and sensitivity (Sens) for each feature. Performances obtained with median frequencies over time(Tdetect 0.91 ± 0.47 s; Sens 0.96) were better than those obtained using the hip angle signal in time domain (Tdetect 1.19 ± 0.27 s; Sens 0.83). Other features did not show significant results. Thus, median frequency over time expected to achieve effective real-time event detection systems, with the aim of a future on-board application concerning detection and prevention measures.

Cortical reactive balance responses to unexpected slippages while walking: a pilot study.

Understanding how the human brain cortex behaves when the dynamical balance is unexpectedly challenged can be useful to enable fall prevention strategies during daily activities. In this respect, we designed and tested a novel methodological approach to early detect modifications of the scalp-level signals when steady walking is perturbed. Four young adults were asked to manage unexpected bilateral slippages while steadily walking at their self-selected speed. Lower limb kinematics, electromyographic (EMG) and electroencephalographic (EEG; 13 channels from motor and sensory-motor cortex areas) signals were synchronously recorded. EMG signals from Vastus Medialis (both sides) were used to trigger the analysis of the EEG before and after the perturbation onset. Cortical activity was then assessed and compared pre vs. post perturbation. Specifically, for each gait cycle, the rate of variation of the EEG power spectrum density, named m, was used to describe the cortical responsiveness in five bands of interests: ϑ (4-7 Hz), α (8-12 Hz), β I, β II, β III rhythms (13-15, 15-20, 18-28 Hz). Results revealed a sharp increment of m early after the onset of the perturbation (perturbed step) compared to steady locomotion, for all rhythms. This cortical behavior disappeared during the recovery step. This study promisingly supports the evidence that the proposed approach can distinguish between steady walking and early reactive balance recovery, paving the way for the EEG-based monitoring of the fall risk during daily activities.

A lumped parameter model of endoplasm flow in Physarum polycephalum explains migration and polarization-induced asymmetry during the onset of locomotion.

The plasmodial slime mold Physarum polycephalum exhibits strong, periodic flow of cytoplasm through the veins of its network. In the special case of mesoplasmodia, a newly described starvation-induced, shape-constant morphotype, this periodic endoplasm streaming is the basis of locomotion. Furthermore, we presume that cytoplasm flow is also involved in signal transmission and signal processing. Mesoplasmodia motility resembles amoeboid locomotion. In contrast to other amoebae, however, mesoplasmodia move without extending pseudopods and retain a coherent, fan-shaped morphology throughout their steady locomotion. Attaining sizes of up to 2 mm2, mesoplasmodia are also much bigger than other amoebae. We characterize this particular type of locomotion and identify patterns of movement. By using the analogy between pulsatile fluid flow through a network of elastic tubes and electrical circuits, we build a lumped model that explains observed fluid flow patterns. Essentially, the mesoplasmodium acts as a low-pass filter, permitting only low-frequency oscillations to propagate from back to front. This frequency selection serves to optimize flow and reduces power dissipation. Furthermore, we introduce a distributed element into the lumped model to explain cell polarization during the onset of chemotaxis: Biochemical cues (internal or external) lead to a local softening of the actin cortex, which in turn causes an increased flow of cytoplasm into that area and, thus, a net forward movement. We conclude that the internal actin-enclosed vein network gives the slime mold a high measure of control over fluid transport, especially by softening or hardening, which in turn leads to polarization and net movement.

Effects of repeated waist-pull perturbations on gait stability in subjects with cerebellar ataxia

BackgroundDamage to the cerebellum can affect neural structures involved in locomotion, causing gait and balance disorders. However, the integrity of cerebellum does not seem to be critical in managing sudden and unexpected environmental changes such as disturbances during walking. The cerebellum also plays a functional role in motor learning. Only a few effective therapies exist for individuals with cerebellar ataxia. With these in mind, we aimed at investigating: (1) corrective response of participants with cerebellar ataxia (CA) to unexpected gait perturbations; and (2) the effectiveness of a perturbation-based training to improve their dynamic stability during balance recovery responses and steady walking. Specifically, we hypothesized that: (1) CA group can show a corrective behavior similar to that of a healthy control group; (2) the exposure to a perturbation-based treatment can exploit residual learning capability, thus improving their dynamic stability during balance recovery responses and steady locomotion.MethodsTen participants with cerebellar ataxia and eight age-matched healthy adults were exposed to a single perturbation-based training session. The Active Tethered Pelvic Assist Device applied unexpected waist-pull perturbations while participants walked on a treadmill. Spatio-temporal parameters and dynamic stability were determined during corrective responses and steady locomotion, before and after the training. The ANalysis Of VAriance was the main statistical test used to assess the effects of group (healthy vs CA) and training (baseline vs post) on spatio-temporal parameters of the gait and margin of stability.ResultsData analysis revealed that individuals with cerebellar ataxia behaved differently from healthy volunteers: (1) they retained a wider base of support during corrective responses and steady gait both before and after the training; (2) due to the training, patients improved their anterior-posterior margin of stability during steady walking only.ConclusionsOur results revealed that participants with cerebellar ataxia could still rely on their learning capability to modify the gait towards a safer behavior. However, they could not take advantage from their residual learning capability while managing sudden and unexpected perturbations. Accordingly, the proposed training paradigm can be considered as a promising approach to improve balance control during steady walking in these individuals.

History-dependent perturbation response in limb muscle.

Muscle mediates movement but movement is typically unsteady and perturbed. Muscle is known to behave non-linearly and with history-dependent properties during steady locomotion, but the importance of history dependence in mediating muscle function during perturbations remains less clear. To explore the capacity of muscles to mitigate perturbations during locomotion, we constructed a series of perturbations that varied only in kinematic history, keeping instantaneous position, velocity and time from stimulation constant. We found that the response of muscle to a perturbation is profoundly history dependent, varying 4-fold as baseline frequency changes, and dissipating energy equivalent to ∼6times the kinetic energy of all the limbs in 5 ms (nearly 2400 W kg-1). Muscle energy dissipation during a perturbation is predicted primarily by the force at the onset of the perturbation. This relationship holds across different frequencies and timings of stimulation. This history dependence behaves like a viscoelastic memory producing perturbation responses that vary with the frequency of the underlying movement.

Real-Time On-Board Recognition of Continuous Locomotion Modes for Amputees With Robotic Transtibial Prostheses.

Human intent recognition is important to the control of robotic prosthesis. In this paper, we propose a multi-level real-time on-board system to recognize continuous locomotion modes. A cascaded classification strategy is designed for the recognition of six steady locomotion modes and 10 transitions. On-board signals of the robotic prosthesis include two inertial measurement units and one load cell. Three transtibial amputees are recruited in the experiments. The prediction decision time of the real-time on-board cascaded classification system is about 3.3 ms, which is enough short compared with the sliding window increment 10 ms. It is easy to recognize the standing and ambulation in the first-level classification with a 99.86% accuracy by quadratic discriminant analysis (QDA) classifier. In the second-level classification, threshold method is adopted to divide one stride into swing and stance phases. In swing phase, five steady modes are recognized with a total accuracy of 96.40%. In stance phase, all these five steady modes are recognized with a total accuracy of 91.21%. The average recognition accuracy of the three subjects is 93.21% by QDA classifier. Besides, for transitions, the proposed system could recognize all transitions rightly. The designed system is feasible and effective to realize real-time on-board recognition of continuous locomotion modes, which is promising for the further control of the prosthesis.