Kenzo Akazawa

Static torque-angle relation of human elbow joint estimated with artificial neural network technique

Static relations between elbow joint angle and torque at constant muscle activity in normal volunteers were investigated with the aid of an artificial neural network technique. A subject sat on a chair and moved his upper- and forearm in a horizontal plane at the height of his shoulder. The subject was instructed to maintain the elbow joint at a pre-determined angle. The wrist was then pulled to extend the elbow joint by the gravitational force of a weight hanging from a pulley. Integrated electromyograms (IEMGs), elbow and shoulder joint angles and elbow joint torque were measured. Then the relation among IEMGs, joint angles and torque was modeled with the aid of the artificial neural network, where IEMGs and joint angles were the inputs and torque was the output. After back propagation learning, we presented various combinations of IEMGs, shoulder and elbow joint angles to the model and estimated the elbow joint torque to obtain the torque-angle relation for constant muscle activation. The elbow joint torque increased and then decreased with extension of the elbow joint. This suggests that if the forearm is displaced from an equilibrium point, the torque angle relation would not act like a simple spring. In a view of the musculoskeletal structure of the elbow joint, the relation between the elbow joint angle and the moment arm of the elbow flexor muscles seems to have a dominant effect on the torque-angle relation.

Biomimetic EMG-prosthesis-hand

The purpose of this study is to develop a myoelectric prosthetic hand that simulates fundamental dynamic properties of the neuromuscular control system of human hand. In particular, this prosthetic hand mimics the property that both muscle viscoelasticity and gain of the stretch reflex vary linearly with the activity of muscle. Those nonlinear properties of the neuromuscular control system were realized by using a position control system of the finger movement, force feedback and variable gain which was modulated by amplitude of rectified and smoothed EMG signals. The prosthetic hand consisted of a processing unit of surface EMG signals, a digital servo system of DC motor and a mechanical hand of one-degree-of-freedom with three fingers. Usefulness of the developed prosthetic hand was shown in myoelectric control experiments with an amputated subject. Both the finger-angle and the stiffness of the hand were voluntarily controlled with EMG signals and soft objects were grasped smoothly.

Feature extraction using evolutionary computation

We propose a method of feature extraction to improve the performance of pattern recognition. The extracted features are assumed to be a polynomial expression of the original patterns. The polynomial expressions are searched by the genetic programming. In order to evaluate the effectiveness of the proposed method, we apply k nearest neighbor classifier as the classification algorithm. Experiments were performed for two artificial tasks and an acoustic diagnosis for compressors as the real world task. From these results, we confirmed that the proposed method was effective for the feature extraction.

Application of independent component analysis to feature extraction of speech

We describe what characteristics an independent component analysis can extract from Japanese continuous speech. Speech data was selected from ATR database uttered by a female speaker. The data was recorded at 20 kHz sampling frequency and was pre-processed with a whitening filter. The learning algorithm of a network was an information-maximization approach proposed by Bell and Sejnowski (1995). After the learning, most of the basis functions that are columns of a mixing matrix were localized in both time and frequency. Furthermore, we confirmed that there were some basis functions to extract the acoustic feature such as the pitch and the formant of each vowel.

Development of biomimetic prosthetic hand controlled by electromyogram

The purpose of this study was to develop a new type of myoelectrically-controlled biomimetic prosthetic hand (biomimetic Kobe Hand) which had almost the same dynamics as those of the neuromuscular control system of finger muscles, in particular, including mechanical properties of the muscle and of the stretch reflex. One of the characteristic features of the neuromuscular control system in man was a decrease in the compliance around the joint with increasing activities of the muscle. The Kobe hand consisted of a processing unit of surface EMG signals, a digital servo system of DC motor and a mechanical hand of one-degree-of-freedom with three fingers. Those nonlinear properties of the neuromuscular control system were realized by using a position control system of the finger movement, force feedback and variable gain which was modulated by amplitude of IEMGs (rectified and smoothed electromyograms). EMG signals picked up from a pair of antagonistic muscles of the forearm were used as control signals. Usefulness of the biomimetic Kobe hand was showed by executing myoelectric control experiments with both healthy and amputated subjects.

Neuromuscular control system and hardware models

Both the mechanism of neuromuscular control system and its hardware models are outlined from the control point of view, focusing on the regulation of mechanical impedance. First, modulations of the mechanical impedance (both the muscle stiffness and the stretch reflex) experimentally observed in a walking cat are demonstrated. Second, the basic idea of hardware models which could simulate the dynamics of the neuromuscular control system is explained. Finally, a myoelectric prosthetic hand is depicted as an example of application of the hardware model.

Adaptive data compression of ambulatory ECG using multi templates

Proposes a new adaptive method of data compression for digital ambulatory electrocardiograms (ECGs), considering the diagnostic significance of each segment of the ECG. The R-wave is detected, followed by multi-template matching of the detected beat and judgment of the noise level; the templates are successively created during processing. The residual signal (the difference between the original ECG and the best-fit template) is approximated with the FAN data compression method SAPA2 (Scan-Along Polygonal Approximation) and then encoded. The error threshold of FAN is decreased during the P-wave segments and increased during the noise segments; the maximum error of the reconstructed signal at each time is known. This method is applied to ECGs of the AHA (American Heart Association) database and its usefulness is indicated; e.g. the bit rate is approximately 400 bps at 8% PRD (percent RMS difference) and 200 bps at 15% PRD.<<ETX>>

Cybernetic actuators. Muscle mechanics and hardware model

Mechanical properties of the skeletal muscle are outlined from the engineering point of view, focusing on the modulation of the viscoelasticity. It is shown that the dynamic property of the muscle varies with the activation level of voluntary contraction in man. Basic ideas of the hardware model (artificial cybernetic actuator),whose mechanical property is similar to those of the contracting muscle, are explained. Finally, a myoelectric prosthetic hand is depicted as an example of the application of the hardware model.<<ETX>>

Neural Network Model for Muscle Force Control Based on the Size Principle and Recurrent Inhibition of Renshaw Cells

A neural network model for muscle force control was constructed. The model contained a single motor-cortex output cell, the actual number of a motoneurons found in human muscles, Renshaw cells and muscle units. The size of the motor units (motoneurons and muscle units) was distributed as the human brachialis muscle, the extensor digitorum muscle and the first dorsal interosseous muscle. The relationship between the model’s muscle force and the firing rate of a motoneurons was investigated. The relationship depended on the absolute refractory time of a motoneurons, RIPSP by Renshaw cells and the firing pattern of Renshaw cells. When these parameters were selected appropriately, the model showed a relationship similar to that observed in isometric contraction of human skeletal muscles. The size distribution of the motor units had a dominant effects on the relationship.

Modulation and Adaptability of Mechanical Properties of Mammalian Skeletal Muscle

Basic mechanical properties of the isolated skeletal muscle fiber are outhned in this chapter. Behavior of the motor units and modulation of the mechanical properties in voluntary contraction are then explained. Finally, findings concerning the morphological and metabolic changes resulting from nerve-induced activities or exercise training are described. (1) Viscoelasticity of the whole muscle becomes greater almost linearly with an increase in force of the voluntary contraction. This is attributed to increase in the number of active muscle fibers and in their activities; that is, when the force of the muscle is increased, the motor units are recruited in order of increasing size and their firing rates are increased. (2) Exercise training is accompanied by an increase in fiber size of the muscle (hypertrophy), and possibly by an increase in fiber number (hyperplasia). The transformation of the muscle phenotype occurs in animal experiments; fast muscle fibers that are made to perform a more tonic activity become slower-contracting.