Medical Engineering & Physics 21 (1999) 87���94 www.elsevier.com/locate/medengphy A practical gait analysis system using gyroscopes Kaiyu Tong, Malcolm H. Granat * Bioengineering Unit, Wolfson Centre, University of Strathclyde, Glasgow G4 0NW, UK Received 28 October 1998 received in revised form 4 February 1999 accepted 3 March 1999 Abstract This study investigated the possibility of using uni-axial gyroscopes to develop a simple portable gait analysis system. Gyroscopes were attached on the skin surface of the shank and thigh segments and the angular velocity for each segment was recorded in each segment. Segment inclinations and knee angle were derived from segment angular velocities. The angular signals from a motion analysis system were used to evaluate the angular velocities and the derived signals from the gyroscopes. There was a good correlation between these signals. When performing a turn the signals of segment inclination and knee angle drifted. Two methods were used to solve this: automatically resetting the system to re-initialise the angle in each gait cycle, and high-pass filtering. They both successfully corrected this drift. A single gyroscope on the shank segment could provide information on segment inclination range, cadence, number of steps, and an estimation of stride length and walking speed. ��� 1999 IPEM. Published by Elsevier Science Ltd. All rights reserved. Keywords: Ambulatory monitor Gait analysis Gyroscope Sensors 1. Introduction Gait analysis has become a widely used clinical tool, and an increasing number of physical therapists and doc- tors are choosing suitable treatments for their patients based on the information from kinematic and kinetic data [1]. Kinematic data are also increasingly being used for the control of neural prostheses or functional electrical stimulation (FES) [2,3]. A complete gait analysis system uses an optical motion analysis system for kinematic data combined with force platforms for kinetic data. These systems are expensive, require a large space and cannot be used out- side a laboratory environment. The capture volume is also limited to a few gait cycles. Therefore there has been much activity in trying to find alternative solutions for capturing gait information over a larger distance and outside a laboratory environment [3���10]. During the last decade, many sensors have been developed for industrial, robotics, aerospace and biom- edical measurements using the continuously advancing circuit technology. These sensors are becoming more compact in size and lower in cost [4,5]. For gait analysis, * Corresponding author. 1350-4533/99/$ - see front matter. ��� 1999 IPEM. Published by Elsevier Science Ltd. All rights reserved. PII: S 1350- 45 33 (99)00 03 0- 2 electrogoniometers, accelerometers, inclinometers and force sensitive resistors (FSR) can be used to measure joint angles, linear acceleration, tilt angle relative to gravity and times of foot contact, respectively. A port- able kinematic gait analysis system can be built using these small sensors. If a number of these sensors are used simultaneously, the system can become cumbersome and difficult to don/doff. A practical system must be small and easy to apply, and provide enough relevant infor- mation. If the position and the orientation of each body seg- ment is known, then it is possible to calculate all the kinematic data. In the aerospace industry, gyroscopes and accelerometers are widely used to provide infor- mation on position and orientation. It is theoretically possible to use the same techniques for gait analysis. The signals from accelerometers and gyroscopes are acceler- ation and angular velocity, respectively. The raw signals from accelerometers and gyroscopes have been used to quantify human daily activities [6���10]. Joint angles are commonly used in gait analysis and can be derived by the integration of angular acceleration or angular velo- city. However, data obtained from integration can be dis- torted by offsets or any drifts. To find the angular acceleration using accelerometers, a pair of accelerometers fixed on a rigid object is

88 K. Tong, M.H. Granat / Medical Engineering & Physics 21 (1999) 87���94 required. In order to eliminate any drift during inte- gration, Morris (1973) [11] identified the beginning and the end of the walking cycles, and made the signal at the beginning and the end of the cycle equal. Willemsen et al. (1990) [12] developed a technique to find the joint angle without the need for integration, which used four accelerometers on each segment. The system used two metal bars with eight accelerometers for measuring a sin- gle joint angle. They also used a simplified version of their technique for the control of an FES system. Four accelerometers on a metal bar were used to calculate the joint acceleration, and different phases of gait could be detected for FES control without the need for the angular information [2]. Inclinometers have also been evaluated for use in con- trolling FES systems [3]. Inclinometers detect inertial forces. During the stance phase, when the angular accel- eration is nearly zero, the inertial force is principally due to gravity, and the segment inclination can then be calcu- lated. During the swing phase the angular acceleration affects the measurement and therefore inclination cannot be accurately calculated. Another promising alternative is to use gyroscopes directly to measure the angular velocity without the sig- nal being affected by gravity or any linear acceleration. Gyroscopes can therefore theoretically be used to calcu- late the segment inclination and the relative joint angle. During walking the movement of the lower body seg- ments occurs mainly in the sagittal plane, so only single uni-axial gyroscopes would be required on each seg- ment. Heyn et al. (1996) [13] had showed that shank incli- nation could be measured with eight accelerometers and two gyroscopes fixed on two rigid metal plates. This experimental protocol did not include any turning which could be expected to effect the inclination. They also found that using these metal plates was cumbersome. The aim of this study was to investigate the possibility of using uni-axial gyroscopes to design a practical gait analysis system. The gyroscopes would be fixed directly to the skin making the system easy to apply and reducing subject encumbrance. The first objective was to evaluate the angular signals and derived signals from gyroscopes and compare these with data from the motion analysis system (Vicon). The second objective was to evaluate the problem of drifting during turning and to investigate solutions to this problem. 2. Methods The dynamic performance of the uni-axial gyroscopes was evaluated from data collected while a subject walked in a straight line in a gait laboratory. The prin- ciple of operation of the gyroscope is the measurement of the Coriolis acceleration of a vibrating device. It con- sists of a triangular prism made of a special substance called ���Elinvar���. If the prism is rotated about its sense axis the signal is proportional to the angular velocity. The gyroscope used was ENC-05EA (Murata, Japan), and the dimensions of this sensor were 20 3 7.2 3 10 mm. The gyroscope was fixed on both the shank and the thigh segments using a strap (Fig. 1). The sensing axis was along the medial���lateral direction so that the angle in the sagittal plane could be measured. Two subjects were used, an incomplete spinal cord injured (SCI) sub- ject and an unimpaired subject. A motion analysis sys- tem (Vicon) was used to evaluate segment inclinations, segment angular velocities and knee angle using retro- reflective markers attached to anatomical positions on the thigh and shank segments. Four FSRs were placed underneath the foot (one under each of the big toe, first metatarsal, fifth metatarsal and heel) and signals from these FSRs were used to detect different gait phases. All the signals were synchronously recorded using the Vicon system at a 50 Hz sampling frequency. In the first 5 s of each experimental trial, the subject stood still in upright position to initialise the inclination angle and gyroscope offset. Then the subject walked at his preferred speed along the walkpath. Three sets of experiments were conducted to analyse the performance of the gyroscopes. The stride length, the gait cycle time and the speed of each walking session were calculated from the Vicon data. In the first experiment, the signals from gyroscopes attached on two different positions of the shank on the unimpaired subject were compared. In the second experiment the signals from gyroscopes were compared Fig. 1. Strap system with gyroscope. The gyroscope was attached to the strap. During walking, the strap was tied around the limb to secure the position of the gyroscope.