Nayan M. Kakoty

Classification of Grasp Types through Wavelet Decomposition of EMG Signals

In this paper, we present a methodology to classify grasp types based on two channel forearm electromyogram signals. Six grasp types are identified. Classification is through support vector machine using radial basis function kernel based on sum of wavelet decomposition coefficients of the electromyogram signals. In a study involving six subjects, we achieved an average recognition rate of 86%; better than that reported in the liteature.

Recognition of grasp types through principal components of DWT based EMG features

With the advancement in machine learning and signal processing techniques, electromyogram (EMG) signals have increasingly gained importance in man-machine interaction. Multifingered hand prostheses using surface EMG for control has appeared in the market. However, EMG based control is still rudimentary, being limited to a few hand postures based on higher number of EMG channels. Moreover, control is non-intuitive, in the sense that the user is required to learn to associate muscle remnants actions to unrelated posture of the prosthesis. Herein lies the promise of a low channel EMG based grasp classification architecture for development of an embedded intelligent prosthetic controller. This paper reports classification of six grasp types used during 70% of daily living activities based on two channel forearm EMG. A feature vector through principal component analysis of discrete wavelet transform coefficients based features of the EMG signal is derived. Classification is through radial basis function kernel based support vector machine following preprocessing and maximum voluntary contraction normalization of EMG signals. 10-fold cross validation is done. We have achieved an average recognition rate of 97.5%.

A Two Layered Control Architecture for Prosthetic Grasping

Abstract This paper presents a two layered control architecture - Superior hand control (SHC) followed by Local hand control (LHC) for an extreme upper limb prosthesis. The control architecture is for executing grasping operations involved in 70% of daily living activities. Forearm electromyogram actuated SHC is for recognition of user’s intended grasp. LHC control the fingers to be actuated for the recognized grasp. The finger actuation is controlled through a proportionalintegral- derivative controller customized with fingertip force sensor. LHC controls joint angles and velocities of the fingers in the prosthetic hand. Fingers in the prosthetic hand emulate the dynamic constraints of human hand fingers. The joint angle trajectories and velocity profiles of the prosthetic hand finger are in close approximation to those of the human finger

Development of an electromyographic controlled biomimetic prosthetic hand

This paper presents the development of an electromyogram controlled extreme upper limb prosthetic hand prototype following a biomimetic approach. The biomimetic approach is followed to harmonise both physical and functional aspects of the human hand. The prototype with tendon driven fingers possesses 15 degrees of freedom. It can perform grasping operations involved during 70% of daily living activities. The control is in two stages: superior hand control SHC and local hand control LHC. The SHC involves a grasp recognition module responsible for recognition of the grasp type indented by the user. The LHC involves a translation module responsible for emulating the identified grasp type into the prototype. The results obtained for finger joint trajectories during a precision grasp through the two stage control architecture in conformity to the human hand finger.

Towards electromyogram-based grasps classification

This paper details a strategy of discriminating grasp types using surface electromyogram (EMG) signals, which has the potential to be applied for controlling extreme upper limb prosthesis. We have investigated the recognition of six grasp types used during 70% of daily living activities based on two-channel EMG. A grasp classification architecture and feature set have been proposed through the iterative development of the feature set as well as the classifier. Three different classifiers and a variety of features have been explored. From the experimental results, we have hypothesised that continuous wavelet transform function coefficients of the EMG signals having entropy values close to the entropy values of preprocessed EMG signals possess maximum informations about the grasp types. Further, sum of discrete wavelet transform coefficients of EMG signals has been established as a primal feature for grasp classification.

Local Hand Control for Tezpur University Bionic Hand Grasping

Tezpur University (TU) Bionic Hand is a biomimetic extreme upper limb prosthesis. The Hand is intended to emulate grasping operations involved during 70% of daily living activities and have been developed using a biomimetic approach. This paper focus on the development of a local hand control for grasping by TU Bionic Hand. Grasp primitives: finger joint angular positions and joint torques are derived through kinematic and dynamic analysis. TU Bionic Hand emulates the grasp types following the dynamic constraints of human hand. The joint angle trajectories and velocity profiles of the Hand finger are in close approximation to those of the human finger.

A biomimetic similarity index for prosthetic hands

Extreme upper limb prosthesis is a well researched problem. There are a number of research prototypes and a few sophisticated commercially launched variants. For a wider acceptance among amputees, prosthetic hands need to be anthropomorphic i.e. replicate the human hand in form and function. However, it is often difficult to compare and rank prosthetic hands on the extent of their being anthropomorphic. The focus of this paper is to evolve a framework for quantification of anthropomorphism for prosthetic hands. Using Formal Concept analysis, a formal context of anthropomorphism is constructed. Within such a context, an index expressing similarity between the prosthetic and the human hand is derived. Following on the lines of the functional similarity metric for design-by-analogy put forward by McAdams and Wood, a formalism to compare different prosthetic hands to a human hand based on a function-vector for each prosthesis expressed in terms of a set of functional and geometric characteristics is presented. Function-vector is characterized within a formal context of anthropomorphism. The Biomimetic Similarity Index (BSI) so computed reflects extent of anthropomorphism and allows a quantitative comparison of different prosthetic hands. Biomimetic design leads to higher anthropomorphism and should result in a higher BSI. We explore the case of TU Bionic Hand and compare the BSI for five different prosthetic hands.

A Review of Underwater Robotics, Navigation, Sensing Techniques and Applications

The focus of this paper is to review the history of underwater robotics, advances in underwater robot navigation and sensing techniques, and an emphasis towards its applications. Following an introduction, the paper reviews development of the underwater robots since the mid 19th century to recent times. Advancements in navigation and sensing techniques for underwater robotics, and their applications in seafloor mapping and seismic monitoring of underwater oil fields were reviewed. Recent navigation and sensing techniques in underwater robotics has enabled their applications in visual imaging of sea beds, detection of geological samples, seismic monitoring of underwater oil fields and the like. This paper provides a recent review of underwater robotics in terms of history, navigation and sensing techniques, and their applications in seafloor mapping and seismic monitoring of underwater oil fields.

Development of Cost Effective EMG Controlled Three Fingered Robotic Hand

One of the major problems confronted in the development of robotic hand prostheses is the achievement of electromyogram (EMG) based control in terms of grasping adaptability. Moreover prosthetic hands that have appeared in the market are out of the reach of common people because of high cost. In this paper, we report the development and control of a three fingered robotic hand motivated by these facts. The developed hand possess six degrees of freedom and can grasp oval, cuboid, circular and cylindrical objects with self adaptability. A grasp planner based on EMG signals commands the controller for actuating the hand opening and closing. In the grasp planner, a fuzzy classifier recognizes shoulder abduction and adduction based on the root mean square value of EMG. The control is through a proportional controller customized with position and touch sensors. The envisaged cost of the proposed hand would be appreciably low compared to most of the hands available in the market.

Wavelet Selection for EMG Based Grasp Recognition through CWT

This paper details a strategy of discriminating grasp types using surface electromyogram (EMG) signals, which has the potential to be applied for controlling advanced prosthesis for extreme upper limb amputees. We have investigated the classification of six basic grasp types used during 70% of daily living activities. The feature vector for EMG based grasp recognition was derived using continuous wavelet transform (CWT). The proper wavelet basis function was selected through computation of entropy of the preprocessed EMG signals and wavelet transform coefficients of six different wavelet families: Gaussian, Daubechies, Morlet, Mayer, Mexicanhat and Symlet. Based on this, Gaussian wavelet function has been concluded to be possessing maximum informations about grasp types. Experimental results have validated our hypothesis that the CWT coefficients having entropy values close to the entropy of preprocessed EMG signals possesses maximum informations about the grasp types. Classification was through one vs. all multi-class support vector machine with linear kernel following preprocessing and maximum voluntary contraction normalization of EMG signals. We have achieved an average recognition rate of 80% (using the Gaussian wavelet function) cross validated through 10-fold cross validation.