Literature Review. Humans navigate by using visual cues to perceive depth from two-dimensional images,

Transcription

1 Sameer Kamath Literature Review Visual Depth Perception Humans navigate by using visual cues to perceive depth from two-dimensional images, which allows them to navigate. These cues can be divided into oculomotor, monocular, and binocular cues. Binocular cues require both eyes to be open, whereas monocular cues require only one eye to be open. Oculomotor cues rely upon the tensions in eye muscles and the positions of the eyes. As the distance between an object and a person s eyes varies, the eyes move inward toward the nose or outward away the nose. This movement causes the eye muscles to change tension. The relationship between the tension and positon of the eyes provides depth information to the brain. Additionally, monocular cues such as the sizes of familiar objects, shadows, texture gradient, linear perspective, motion parallax, and occlusion further enhance depth perception. Occlusion occurs when an object is covered by other objects, causing the human brain to perceive the covered object as being farther away. Motion parallax causes nearby objects to appear to move through the field of vision faster. Because of motion parallax, objects that are farther away appear to move more slowly. Figure 1. Familiar Size Cues. ( Perceiving Depth, n.d.) Furthermore, the brain uses binocular cues such as stereopsis, the disparity between images the two eyes produce, to calculate depth from the correspondence of the points in the two

2 images ( Perceiving Depth, n.d.). All these complex cues simultaneously work together to provide sighted humans with detailed three dimensional perceptions of their surroundings. Blindness There are an estimated 285 million visually-impaired people, 39 million of which are blind ( Visual Impairment, 2014). Visual impairment indicates that a person has vision 20/40 or worse in his or her better eye when the eye is corrected ( Blindness and Vision Impairment, 2011). A person with 20/40 vision can read a line of letters at 20 feet that a normal person can read at 40 feet ( What Does 20/20 Vision Mean?, n.d.). A person is legally blind if his or her corrected vision in his or her best eye is 20/200. A visual field less than or equal to 20 degrees in diameter in a person s best, corrected eye also constitutes legal blindness. ( Blindness and Vision Impairment, 2011). A person s depth perception and ability to navigate can be severely hindered if his or her visual impairment is substantial. Assistive Technology Many devices have been engineered to help blind people navigate. The white and red cane is one of the simplest of these devices. Blind people sweep the cane from side to side with the tip of the cane hovering above the ground (Edison, 2014). The canes provide people with information about obstructing objects and changes in surface and elevation (Schellingerhout, Bongers, Grinsven, Smitsman, Galen, 2001). Figure 2. Obstacle Detection with a Red and White Cane. (Edison, 2014)

3 Unfortunately, according to several cane-users, canes are inaccurate and tend to miss obstacles (Schellingerhout et al., 2001). These flaws can pose a significant threat to the safety and peace of mind of blind people when navigating. Also, the majority of cane-users hold the cane at angles greater than 35 to the ground, which is far too high to adequately detect many obstacles on the ground (Schellingerhout et al., 2001). Approximately 2% of blind people use trained guide dogs. Guide dogs follow commands and lead their owners around obstacles ( Travel Tools, n.d.). However, guide dogs can cost upwards of $42,000, which makes them unaffordable for many people ( FAQ, 2016). The lack of decent assistive equipment for blind people has lead researchers to utilize more advanced technology, such as sensors and microcontrollers, to engineer complex electronic aids ( National Research Council, 1986). Infrared and ultrasonic sensors are two common types of distance sensors that are widely available. Infrared Sensors Infrared sensors emit a beam of infrared light, and the light is reflected back to a receiver if an obstacle is present. This returning beam creates a triangle between the emitter, obstacle, and receiver. The light is transmitted from the receiver of the sensor to a device called a chargecoupled (CCD) array. The sensor then uses the CCD array to calculate the base angle of the triangular beam, as shown in Figure 3. The base angle is used to calculate the distance to the obstacle (Al-Fahoum, Al-Hmoud, Al-Fraihat, 2013).

4 Points of Emission Figure 3. Infrared Sensor Triangulation. (Al-Fahoum et al., 2013) Unfortunately, infrared sensors, when compared to ultrasonic sensors, are terrible at detecting objects that are not perpendicular to the sensor (Loven, 2016). Furthermore, reflecting infrared light is a tricky process; sunlight can cancel out the infrared beams, and non-reflective surfaces render these sensors useless. (Kanwal, Bostanci, Currie, Clark, 2015). Ultrasonic Sensors The ultrasonic sensor is another common type of sensor that can be used to find distances to objects. The transmitting piezoelectric transducer of an ultrasonic sensor emits a short ultrasonic pulse. After the pulse reflects off an object, the receiving transducer times how long it takes for the pulse to be received. This duration, along with the speed of sound in air, 343 m/s, is used to calculate the distance to the target object. For optimum detection, the obstacle must be relatively flat and capable of reflecting sound properly. It is very important that the sensor face is parallel to the object. If the face of the sensor is not parallel to the surface of the object, the readings can be inaccurate due to the cone-shaped spread of the sound waves ( Ultrasonic Acoustic Sensing, n.d.).

5 Figure 4. Inaccuracies with Ultrasonic Sensors ( Ultrasonic Acoustic Sensing, n.d.) As the energy of the sound beam spreads, beam divergence occurs, as shown in Figure 5. The greatest sound pressure comes from the centerline of the transducer. Therefore, objects directly in front of the transmitting transducer will send the strongest echoes to the receiving transducer. Figure 5. Beam Divergence from a Transducer ( Transducer, n.d.) Sound is a longitudinal pressure wave that moves through a medium by transferring energy between particles. An ideal sound wave would be a perfectly straight, cylindrical shape. However, in the real world, beam divergence occurs because the particles in the air do not perfectly transfer all their energy to the particles directly in front of them as the sound wave propagates. Beam spread is significantly affected by frequency and transducer diameter. Higher frequencies and larger transducers correspond to less beam spread ( Transducer, n.d.).

6 Arduino Mega Arduino is an open source microcontroller platform for building electronic devices, such as travel aids for the blind. Arduino boards can be programmed using the Arduino integrated development environment (IDE) (Blum, 2013). At the heart of the Arduino Mega, one of Arduino s largest and most versatile boards, is an ATMega2560 integrated circuit. Some features of the Mega are 54 digital input/output (I/O) pins, 15 of which can be used for pulse width modulation, 16 analog inputs, and one 16 MHz crystal oscillator for clock signaling ( Arduino - ArduinoBoardMega2560, n.d.). Ordinary binary digital pins can output and receive a high voltage of 5 V or low voltage of 0 V. The pulse width modulation pins can control the duty cycle of the digital high and low voltages, allowing for an analog output or input. For example, an LED can operate at half brightness by receiving a 50% duty cycle signal from an Arduino board (Hirzel, n.d.). The Arduino s low price and extraordinary (I/O) capabilities make the platform tremendously useful for controlling and receiving data from actuators and sensors. Figure 6. Analog Output with Pulse Width Modulation. (Hirzel, n.d.)

7 Vibrating Motors Vibration motors are commonly used in electronic travel aids to alert blind people about obstacles ( National Research Council, 1986). The fundamental principle behind a vibrating motor is a net centripetal force. The non-symmetrical mass attached to the shaft of the dc motor causes an unbalanced centripetal force, which in turn causes the entire motor to move. The shaft oscillates extremely fast, resulting in constant displacement of the motor, which creates vibration ( AB-004 : Understanding ERM, n.d.). Figure 7. Pancake Vibration Motor ( Coin Vibration Motors, n.d.) Electronic Travel Aids Many engineers have attempted to create electronic travel aids (ETA) that assist blind people with navigation. These devices usually consist of sensors, often ultrasonic or infrared, which gather information about a blind person s surroundings, and some sort of auditory or tactile system of transmitting the information to the person. Some aids, known as obstacle detectors, detect only the general area and direction of obstacles without collecting information about the nature of the obstacles. The Russell Pathsounder, for example, uses a chest-level ultrasonic sensor to detect obstacles. It triggers an auditory and tactile warning if an obstacle is

8 detected within 6 feet of the person. The Mowat Sensor, another sonar ETA, is a handheld device that consists of an ultrasonic sensor and vibration motors that vibrate with intensity proportional to the distance detected by the sensor. Researchers have also engineered a modified version of the Mowat Sensor that plays musical notes with frequencies proportional to the detected ranges. Unfortunately, none of these outdated devices have achieved much success on the market because they are expensive, uncomfortable and impractical. Furthermore, the devices do not properly transmit enough desired information ( National Research Council, 1986). Neuroscientists have recently attempted to reinvent electronic travel aids by trying to connect the tongue to the brain by using a device called BrainPort V100. The device consists of a video camera that is fixed to a pair of glasses, and an array of electrodes attached to the tongue. The pixels of the image from the camera are converted into electric pulses that are sent to the array of electrodes on the tongue. The identification of objects, shapes, and text were tested. After one year of training, 69% of subjects could pass the object recognition test ( FDA, 2015). Unfortunately, even with funding from the US Department of Defense and Google, the device still costs $10,000 ( FDA approves novel tongue sensor, 2015). Furthermore, the images transmitted to the brain are of extremely low resolution, and a cane is still necessary for depth perception and obstacle detection in 3D space ( BrainPort V100 Vision Aid, 2015). Researchers at Chonbuk National University in Korea recently engineered a device that used ultrasonic sensors and vibration feedback on the palm to detect and inform blind people about objects in their path (Jeong, Yu, 2016). Figures 8 and 9 show how the configuration of ultrasonic sensors on the device detects objects in front of users as well as changes in elevation. Because the two-point discrimination of the palm is 11 mm, the vibration motors were spaced

9 over 20 mm apart from each other so that users could easily differentiate between which motors were vibrating on their palms (Jeong et al., 2016). Figure 8. Spatial Obstacle Detection. (Jeong et al., 2016) Figure 9. Array of Ultrasonic Sensors. (Jeong et al., 2016)

10 Figure 10. Basic Indoor Obstacle Course. (Jeong et al., 2016) Figure 11. Complex Outdoor Obstacle Course. (Jeong et al., 2016) Figures 10 and 11 show the testing the device underwent. The test subjects of the experiment in Figure 10 were blind-folded university students, and obstacles were placed in random locations unknown to the test subjects. The obstacles were successfully avoided 90% of the time. In the experiment shown in Figure 11, the test subjects were completely blind people who normally used red and white canes for navigation. Single obstacles were avoided 96.7% of the time, dual obstacles were avoided 60% of the time, the toll bar was avoided 86.7% of the time, and the hanging obstacle only had a 13.3% avoidance rate. Detection of obstacles was not

11 the main issue; the researchers found that an overwhelming majority of the collisions were results of attempts to maneuver around obstacles. In fact, many subjects reported that the vibration feedback would change erratically during the maneuvers, leading to collisions (Jeong et al., 2016). Unfortunately, most blind people still have few alternatives to the red and white cane due to a lack of affordable and modern electronic travel aids that provide ample obstacle avoidance capabilities. Research Plan Engineering Problem Many blind people use canes to navigate. Unfortunately, canes are inaccurate and often miss obstacles. Guide dogs, a common alternative, are excessively expensive due to training costs. There are no affordable, widespread, modern electronic travel aids on the market that allow blind people to fully grasp the 3D space in front of them and avoid obstacles. Engineering Goal The goal of this project is to engineer a tactile device that will allow blind people to perceive and avoid walls, inclines, and obstacles below the waist. The final device must be able to notify a person quickly enough so that he or she can recognize and completely avoid such obstacles. Design Criteria Low Cost High Obstacle Avoidance High Comfort Development

12 Three rows of three ultrasonic sensors will be positioned vertically on the body. The first row will be at shoulder height. The second row will be at the midsection. The last row will be attached to the shins. The 3 by 3 array of stationary sonar sensors will scan and map out the depth of the region in front of the person and send the data to an Arduino. A 3 by 3 array of Arduino-controlled vibration motors will form a tactile hand-held grid that will be no larger than the size of a palm. The intensity and location of the vibrations on the grid will represent the presence of obstacles in the 3-D space in front of the person. Rows will correspond to obstacle range, columns will correspond to horizontal position, and vibration intensity will correspond to height. Testing To test the device, blind folded people will use it to navigate through an obstacle course. The subjects ability to successfully avoid walls, obstacles below the waist, and sharp inclines will be tested. The percentage of obstacles detected and avoided will indicate the obstacle detection capabilities of the device. A stick similar to a red and white visual impairment cane will also be tested for obstacle detection capability. All test subjects will be asked to rate the comfort of both the cane and the new device. Comfort ratings will be based on whether subjects would feel comfortable using the aid on a regular basis.

13 Works Cited Perceiving Depth and Size. (n.d.). Retrieved November 23, 2016, from Visual impairment and blindness. (2014, August). Retrieved November 23, 2016, from Blindness and Vision Impairment. (2011, February 08). Retrieved November 23, 2016, from html Edison, T. (Producer). (2014, April 1). How Blind People Use a White Cane [Video file]. In Youtube. Retrieved November 23, 2016, from Schellingerhout, R., Bongers, R. M., Grinsven, R. V., Smitsman, A. W., & Galen, G. P. (2001). Improving obstacle detection by redesign of walking canes for blind persons. Ergonomics, 44(5), doi: / Travel Tools and Techniques of People Who are Blind or Who Have Low Vision. (n.d.). Retrieved November 23, 2016, from FAQ. (2016). Retrieved November 23, 2016, from Al-Fahoum, A. S., Al-Hmoud, H. B., & Al-Fraihat, A. A. (2013). A Smart Infrared Microcontroller-Based Blind Guidance System. Hindawi, 2013, 1-7. doi: /2013/ Loven, P. (2016, October 30). STEM Project With Arduinos [ to the author]. Kanwal, N., Bostanci, E., Currie, K., & Clark, A. F. (2015). A Navigation System for the Visually Impaired: A Fusion of Vision and Depth Sensor. Applied Bionics and Biomechanics, 2015, doi: /2015/ Ultrasonic Acoustic Sensing. (n.d.). Retrieved November 23, 2016, from National Research Council (US) Working Group on Mobility Aids for the Visually Impaired and Blind. (1986). THE TECHNOLOGY OF ELECTRONIC TRAVEL AIDS - Electronic Travel AIDS: New Directions for Research - NCBI Bookshelf. Retrieved November 23, 2016, from FDA allows marketing of new device to help the blind process visual signals via their tongues. (2015, June 18). Retrieved November 23, 2016, from FDA approves novel tongue sensor for the blind. (2015, June 19). Retrieved November 23, 2016, from

14 BrainPort V100 Vision Aid [Video file]. (2015, November 30). In Youtube. Retrieved November 23, 2016, from Blum, J. (2013). Exploring Arduino: Tools and techniques for engineering wizardry. Indianapolis, IN: Wiley. Arduino - ArduinoBoardMega2560. (n.d.). Retrieved November 23, 2016, from Hirzel, T. (n.d.). PWM. Retrieved November 23, 2016, from AB-004 : Understanding ERM Vibration Motor Characteristics. (n.d.). Retrieved November 23, 2016, from understanding-erm-vibration-motor-characteristics Jeong, G., & Yu, K. (2016, July 12). Multi-Section Sensing and Vibrotactile Perception for Walking Guide of Visually Impaired Person. Sensors, 16(7). doi: /s Coin Vibration Motors. (n.d.). Retrieved November 27, 2016, from Transducer Beam Spread. (n.d.). Retrieved December 06, 2016, from What Does 20/20 Vision Mean? (n.d.). Retrieved December 19, 2016, from