Madhavi Anugolu

Frequency domain surface EMG sensor fusion for estimating finger forces

Extracting or estimating skeletal hand/finger forces using surface electro myographic (sEMG) signals poses many challenges due to cross-talk, noise, and a temporal and spatially modulated signal characteristics. Normal sEMG measurements are based on single sensor data. In this paper, array sensors are used along with a proposed sensor fusion scheme that result in a simple Multi-Input-Single-Output (MISO) transfer function. Experimental data is used along with system identification to find this MISO system. A Genetic Algorithm (GA) approach is employed to optimize the characteristics of the MISO system. The proposed fusion-based approach is tested experimentally and indicates improvement in finger/hand force estimation.

Surface EMG Array Sensor Based Model Fusion Using Bayesian Approaches for Prosthetic Hands

Traditional electromyopgrahic (EMG) measurements are based on single sensor information. Due to the arrangement of skeletal muscle fibers for hand motions, cross talk is an inherent problem when inferring motion/force potentials from EMG data. This paper studies means of using sensor arrays to infer better motion/force potential for prosthetic hands. In particular, a surface electromyographic (sEMG) sensor array is used to investigate multiple model fusion techniques. This paper provides a comparison between three statistical model selection criteria. The sEMG signals are pre-processed using four filters, Butterworth, Chebyshev type-II, as well as Bayesian filters such as the Exponential and Half-Gaussian filter. Output Error (OE) models were extracted from sEMG data and hand force data and compared using a Bayesian based fusion model. The four different filters effect were quantified based on the OE models performance in matching the actual measured data. The comparison indicates a preference for using the sensor fusion technique with preprocessed EMG data using the Half-Gaussian Bayesian filter and the Kullback Information Criterion (KIC).Copyright

Implementation of sEMG-based real-time embedded adaptive finger force control for a prosthetic hand

This paper presents surface electromyographic (sEMG)-based, real-time Model Reference Adaptive Control (MRAC) strategy for a prosthetic hand prototype. The proposed design is capable of decoding the prerecorded sEMG signal as well as the sensory force feedback from the sensors to control the force of the prosthetic hand prototype using a PIC 32MX360F512L microcontroller. The input sEMG signal is preprocessed using a Half-Gaussian filter and fed to a fusion based Multiple Input Single Output (MISO) skeletal muscle force model. This MISO system provides the estimated finger forces to be produced as input to the prosthetic hand prototype. A simple MRAC method along with a two stage embedded design is used for the force control of the prosthetic hand. The sensed force at the fingertip is fed back to the controller for real-time operation. The data is transmitted to the computer through an universal asynchronous receiver/transmitter (UART) interface of the proposed embedded design. Results show good performance in controlling the finger force as well as shortcomings of the mechanical design of the prosthetic hand prototype to be addressed in future.

A hybrid adaptive data fusion with linear and nonlinear models for skeletal muscle force estimation

Position and force control are two critical aspects of prosthetic control. Surface electromyographic (sEMG) signals can be used for skeletal muscle force estimation. In this paper, skeletal muscle is considered as a system and System Identification (SI) is used to model sEMG and skeletal muscle force. The recorded sEMG signal is filtered utilizing optimized nonlinear Half-Gaussian Bayesian filter, and a Chebyshev type-II filter prepares the muscle force signal. The filter optimization is accomplished using Genetic Algorithm (GA). Multi- linear and nonlinear models are obtained with sEMG as input and skeletal muscle force of a human hand as an output. The outputs of these models are fused with a probabilistic Kullback Information Criterion (KIC) for model selection and an adaptive probability of KIC. This approach gives good estimate of the skeletal muscle force.

Real-time embedded frame work for sEMG skeletal muscle force estimation and LQG control algorithms for smart upper extremity prostheses

This paper presents a real-time embedded framework for finger force control of upper extremity prostheses. The proposed system is based on the hypothesis that models describing the finger force and surface Electromyographic (sEMG) signal relationships of healthy subjects can be applied to amputees. An identification/estimation scheme is applied to the collected sEMG and finger force signals in order to infer dynamical models relating the two. A LQG control law is proposed based on this estimation scheme in order to control finger forces of upper extremity prostheses. For the force estimation, filtered sEMG signals from a sensor array and finger force data of a healthy subject are acquired. Real-time estimation and control are implemented using a PIC32MX360F512L microcontroller. In this paper, a novel fusion technique, the Optimized Linear Model Fusion Algorithm (OLMFA) is developed for estimating the skeletal muscle force from the sEMG sensor array in real-time. The sEMG signal is rectified and filtered using a Half-Gaussian filter, and fed to the OLMFA based Multiple Input Single Output (MISO) force model. This MISO system provides the estimated finger force as reference input to the upper extremity prostheses in real-time. A LQG controller is designed to control the finger force of the prostheses utilizing the force estimate from the OLMFA as a reference. Both the OLMFA and the LQG control scheme are prototyped on the embedded framework for testing of the real-time performance. The proposed embedded framework features rate partitioning and UART interface for performance validation and troubleshooting. The OLMFA based force estimation yields a real-time performance of 85.6% mean correlation and 20.4% mean relative error with a standard deviation of ?1.6 and ?1.5 respectively for 18 test subjects k-fold cross validation data. The LQG control algorithm yields a real-time performance of 91.6% mean correlation and 9.2% mean relative error with a standard deviation of ?1.4 and ?1.3 respectively. The novelty lies with the proposed Optimized Linear Model Fusion Algorithm (OLMFA).The novel OLMFA is used as the input to a LQG and MRAC control scheme and is implemented in real-time on a microcontroller.An embedded test bed is proposed to rapid prototype the LQG and MRAC control schemes and validate them on the prosthetic hand prototype.

An adaptive hybrid data fusion based identification of skeletal muscle force with ANFIS and smoothing spline curve fitting

Precise and effective prosthetic control is important for its applicability. Two desired objectives of the prosthetic control are finger position and force control. Variation in skeletal muscle force results in corresponding change of surface electromyographic (sEMG) signals. sEMG signals generated by skeletal muscles are temporal and spatially distributed that result in cross talk between adjacent sEMG signal sensors. To address this issue, an array of nine sEMG sensors is used with a force sensing resistor to capture muscle dynamics in terms of sEMG and skeletal muscle force. sEMG and skeletal muscle force are filtered with a nonlinear Teager-Kaiser Energy (TKE) operator based nonlinear spatial filter and Chebyshev type-II filter respectively. Multiple Takagi-Sugeno-Kang Adaptive Neuro Fuzzy Inference Systems (ANFIS) are obtained using sEMG as input and skeletal muscle force as output. Outputs of these ANFIS systems are fitted with smoothing spline curve fitting. To achieve better estimate of the skeletal muscle force, an adaptive probabilistic Kullback Information Criterion (KIC) for model selection based data fusion algorithm is applied to the smoothing spline curve fitting outputs. Final fusion based output of this approach results in improved skeletal muscle force estimates.

sEMG based fuzzy control strategy with ANFIS path planning for prosthetic hand

This paper presents an intelligent adaptive neurofuzzy inference system (ANFIS) based fuzzy Mamdani controller for a multifingered prosthetic hand. The objective of the controller is to move the finger joint angles along predetermined paths representing a grasping motion. The initiation of the grasping task is evaluated via EMG-entropy data, measured at the forearm of the prosthetic user. In addition to the motion control, the finger force is regulated with a Fuzzy logic controller. Simulation results indicate good performance of the proposed controller. Results show that the outputs follow the hand/finger force and given reference trajectory closely.

Optimal tracking of a sEMG based force model for a prosthetic hand

This paper presents a surface electromyographic (sEMG)-based, optimal control strategy for a prosthetic hand. System Identification (SI) is used to obtain the dynamic relation between the sEMG and the corresponding skeletal muscle force. The input sEMG signal is preprocessed using a Half-Gaussian filter and fed to a fusion-based Multiple Input Single Output (MISO) skeletal muscle force model. This MISO system model provides the estimated finger forces to be produced as input to the prosthetic hand. Optimal tracking method has been applied to track the estimated force profile of the Fusion based sEMG-force model. The simulation results show good agreement between reference force profile and the actual force.

Optimization of Bayesian Filters and Hammerstein-Wiener Models for EMG-Force Signals Using Genetic Algorithm

Processing electromyographic (EMG) signals for force estimation has many unknown variables that can influence the outcome or interpretation of the recorded EMG signal significantly. An array of filtering methods have been proposed over the past few years with the objective to classify motion for use in prosthetic hands. In this paper, we explore the optimal parameter settings of a set of Bayesian based EMG filters with the objective to use the filtered EMG data for system identification. System identification is utilized to establish a relationship between the measured EMG data and the generated force developed by fingers in a human hand. The proposed system identification is based on nonlinear Hammerstein-Wiener models. Optimization is also applied to find the optimal parameter settings for these nonlinear models. Genetic Algorithm (GA) is used to conduct the optimization for both, the optimal parameter settings for the Bayesian filters as well as the Hammerstein-Wiener model. The experimental results and optimization analysis indicate that the optimization can yield significant improvement in data accuracy and interpretation.Copyright

Fusion of spectral models for dynamic modeling of sEMG and skeletal muscle force

In this paper, we present a method of combining spectral models using a Kullback Information Criterion (KIC) data fusion algorithm. Surface Electromyographic (sEMG) signals and their corresponding skeletal muscle force signals are acquired from three sensors and pre-processed using a Half-Gaussian filter and a Chebyshev Type- II filter, respectively. Spectral models - Spectral Analysis (SPA), Empirical Transfer Function Estimate (ETFE), Spectral Analysis with Frequency Dependent Resolution (SPFRD) - are extracted from sEMG signals as input and skeletal muscle force as output signal. These signals are then employed in a System Identification (SI) routine to establish the dynamic models relating the input and output. After the individual models are extracted, the models are fused by a probability based KIC fusion algorithm. The results show that the SPFRD spectral models perform better than SPA and ETFE models in modeling the frequency content of the sEMG/skeletal muscle force data.