Mobile Robot Control with the Electrooculogram Signal João Raminhos1 and João Sanches2 1
Instituto Superior Técnico, Lisbon, Portugal
[email protected]
2
Systems and Robotics Institute, Instituto Superior Técnico, Lisbon Portugal
[email protected]
Abstract—The Electrooculogram (EOG) signal measures the resting potential of the retina, which is assumed to be constant under stationary light conditions. By using two electrodes located at both sides of the eyes and one at the middle it is possible to measure the orientation of the eyes with respect to the skull. This paper describes a control system for a mobile robot based on the EOG signal. The goal is to control the direction and speed of the robot with the eyes using the EOG signal. The acquisition system communicates with the laptop through the audio channel of the sound card and the robot communicates with the laptop through a bluetooth communication link. The software running in the laptop demodulates the AM signal containing the EOG, detect some predefined eye movements and give the corresponding commands to the robot. The whole system and the corresponding hardware and software modules are described and results are presented for illustrative purposes. Index Terms—Robot, opamp, AM, python, Electrooculogram, electrocardiogram, physiological
1. INTRODUCTION The acquisition of several physiological signals, such as the electrocardiogram (ECG), electroencephalogram (EEG) and electromyogram (EMG) is a current practice since several decades, mainly for diagnosis purposes. More recently these signals are being used for control and interface purposes. In this scope, the brain computer interface (BCI) systems have received particular attention in the last decades [1]. These
systems aims at control external devices or processes by finding particular EEG patterns that results from particular simple thoughts. The most common example of physiological signals processing system is the pacemaker. This external device, implanted in the chest is controlled by the electrophysiological cardiac signal (ECG) to generate a trigger signal that regularizes the cardiac rhythm [2]. Other systems have been developed in the last years to help people with some kind of physical disability. Tetraplegic persons are one of those groups of people who most need these kinds of systems. With that goal, some systems have been developed using the EOG signal, like [3], [4] and [5], where the EOG signal is used to control a wheelchair. The system described in [6] makes it possible for the tetraplegic people to control a wheelchair using their own electromyographic (EMG) signals. This system is somewhat different from the one described in this paper. The main difference resides in the control signal that is the EMG in [6] while in the system described here is the EOG. The EMG is noisier than the EOG, because there are several muscles interacting simultaneously and it is not easy to acquire the signal from one particular muscle. Artificial retina, on the contrary, generates electrical signals to be interpreted by the visual cortex. In these systems a video signal is encoded and sent by the optical nerve to the visual area of the cortex [7]. In this paper the authors describe a system
where a mobile platform is controlled using the user’s EOG signal. The system is composed by four main modules: i) EOG signal acquisition and conditioning, ii) transmission, iii) mobile platform communication and control software and iv) mobile platform. The acquisition module acquires, amplifies and filters the EOG signal conveniently for subsequent transmission. It is modulated in amplitude a carrier to transmit the processed signal through an audio channel, e.g., the microphone audio channel of the sound card of a laptop, mobile phone or PDA. The communication with the mobile platform is performed through a bi-directional bluetooth link, making it possible to receive and send information from and to the mobile platform. The software used to demodulate the EOG modulated signal and to detect the events needed for the control is implemented in Python [8] to make it possible to be used in different operating systems and/or different devices. This paper is organized as follows. In section 2 is described the developed system and all of its modules. In section 3 are described the experimental results and section 4 concludes this paper. 2. SYSTEM DESCRIPTION As described above, the developed system has four main modules: the acquisition, the transmission, the processing and the mobile platform, as shown on Fig. 1.
Acquisition System
Transmission
Processing
Fig. 1. Block diagram of the developed system.
Each of the developed modules will be described in detail in the following sections. 2.1. Acquisition The acquisition of the EOG signal is performed
by three electrodes, two of them located at both ends of the sphenoid bone and the third located at the frontal bone, just above the nose, as shown on Fig. 2.
Fig. 2. EOG electrodes placement.
This module is composed by one instrumentation amplifier, which acquires the left and right electrodes, and one operational amplifier, which imposes the reference voltage to the middle electrode. This middle electrode is used to set the reference voltage to a low constant known voltage, defined in the gain loop of the instrumentation amplifier. Because the acquired signal has very low magnitude, in the range of to , it is necessary to amplify it, about 220 times. The overall amplification chain is composed by three stages because it is easier to design stable amplifiers when its gain is small. Therefore, the huge amount of amplification needed in the overall amplification chain is obtained with several (three) medium gain amplification stages. To do that, an amplification chain with four stages (based on rail-to-rail operational amplifier) is used. The first stage is a voltage follower to isolate the acquisition circuit from the amplification chain. Each one of the other three stages is a low gain low-pass filter to reduce the high frequency noise. After the buffering stage, it is placed a bandpass filter so that its frequency response can capture only the EOG signal characteristics, leaving a clean signal to amplify. The first of the remaining amplifying stages has a fixed gain of . The second stage has a maximum gain of , which is controlled by a trimmer in its loop gain. Finally, the last amplifying stage has a gain of , having the peculiarity of introducing an offset to the signal, controlled by a trimmer placed on the
inverting node of the opAmp. The gain and offset controls are important to adjust the output dynamic range of the processed signal to the input dynamic range of the modulator. A block diagram descriptive of the acquisition module described above is shown in Fig. 3 and in Fig. 4 it is shown the amplification chain circuit. Acquisition circuit
Fig. 5. AM modulation diagram.
Fig. 6 displays the modulated square carrier (after high-pass filtered to remove the DC component) and the same carrier obtained in the laptop. The red line represents the continuous carrier signal at the audio channel input and the blue dots represent the correspondent discrete signal before the demodulation process.
Signal conditioning Band-pass filter [0.218, 113.532]Hz
First stage K1=10
Second stage K2