A multisensor electronic traveling aid for the visually impaired

Abstract

BACKGROUND:

Identifying obstacles and potholes in the pathway of the visually impaired have received much attention. While much has been done on the detection of obstacles, in the design of electronic mobility aids, much more is needed to be done on the detection of holes and drop-offs, especially those that are affordable and hands-free. This paper therefore considers the development of a wearable electronic mobility aid.

METHODS:

The developed system is based on the multisensor fusion approach of detection which combined three techniques, namely: a source of laser light, a camera and an ultrasonic sensor. A red line generating laser source is used to project a straight line and this is captured by the camera. The red line is deformed differently on coming in contact with holes or standing obstacles. The pattern of deformation is then extracted for obstacle and pothole recognition. The visibility of laser light is greatly reduced when the scene is extremely illuminated, so this is complemented with edge detection. The edge detection uses edges in the identification of holes and obstacles. This is combined with ultrasonic sensing, so that the presence of obstacles can be differentiated from that of holes. The outcome of detection and the distance of obstacles from the blind are relayed via an audio cue.

REDULTS:

Its evaluation showed better performance compared to the guide cane. It showed a reduction in collision rate by 83.25% and reduction in falling rate by 84.62%. The device received good acceptability from the users.

Keywords

Wearable electronic mobility aid multi-sensor system visually impaired

1. Introduction

It has been observed that out of the 7 billion world population, about 285 million persons suffer from visual impairments, with about 39 million people suffering from total blindness and about 246 million suffering from low vision [1]. Visual impairment has become a global challenge without exceptions to the developed nations [2]. Efforts have been made on several fronts to develop various mobility aids for persons with visual impairments. Devising modern means of navigation for the visually impaired persons has become an important subject of interest of research. Some levels of successes have been attained, however, there are some limitations inherent in the existing solutions that often reduce the effectiveness of their use [3].

The oldest navigation assistance for eye defective persons are based on the use of white cane (i.e. walking cane) and that of the guide dogs. The walking stick was found helpful as it enhances the navigation of the blinds to successfully discharge their regular daily duties. However, there is still some form of limitations with it. One of these limitations is that, a walking stick can only detect below-knee objects [4]. Another is that it can only detect obstacles that are within the immediate radius of the stick and with higher likelihood of the blind getting hit by the obstacle most times because of the short response time.

Trainable domestic animals such as horses, dogs, etc., were adopted as navigation aids for the blinds. However, this method is very costly and requires a lot of training. The use of these methods often suffers some drawbacks as they convey little information to users. They are less user-friendly and often require special skills and training [2]. Recent developments have seen various designs of Electronic Travel Aid (ETA) which are aimed at assisting the visually impaired persons to travel in a safe and independent manner. In [5], a multidimensional walking aid was developed using a network of ultrasonic sensors to detect obstacles in three directions (i.e. left, right and the front of the blinds). A light dependent resistor was used to sense when it is getting dark, and consequently prompt the blind. But the proposed method did not address the problem of identifying potholes on the way. In [6], an approach was proposed in which infra-red sensors were attached to the users shoe, at the front and at the two sides of the shoe, to detect obstacles in three different directions. A colour sensor was employed through which the blind could also detect obstacles. Image processing techniques have also been employed to assist the blind to move freely in an unfamiliar scene. In [7], a smart cane was developed to enhance the navigation of the blind both in indoor and outdoor environments.

An image acquisition device was used to capture the image of the walk path and then sent to a PC computer for processing. Because a PC required for the image processing, it becomes a burden to be carried around. In [8], a smart vision system for the visually impaired was developed for the detection of obstacles both for the interior and the exterior environments. In [9], a user selectable modes smart cane was designed to assist the blind to move within and outside the room. The user can switch the mode depending on the decision either to move within or outside a building. Each of the selectable modes was designed to detect both below-knee and above-knee obstacles. In [10], a wearable belt around the waist with a network of four ultrasonic sensors strategically arranged was proposed. The arrangement was to detect both standing obstacles and holes. A computer was used as the processor which the user would have to carry as he walks. This approach may not allow the blind to move freely as he would have to use his hand to support or carry the device. Two of the ultrasonic sensors were slightly tilted to detect holes. The major concern with this method is that there are possibilities of some holes having a texture that is not very reflective, giving rise to weak reflected waves that may not be easily sensed by the ultrasonic module thereby escaping detection. Though an image processing was later adopted in [10] to aid the detection, it could be very difficult to differentiate edges of potholes from the edges of standing obstacles. Even with this, there is still a major drawback as the focus has mostly been on the avoidance of extruding obstacles which prevents the suitability for general purpose use [3]. Though, much work has been done on the detection of standing obstacles, there is much to be done in holes detection for the blind. It is observed that holes pose the greatest challenge for the blinds [10]. Also, most visually impaired persons still prefer the use of the old walking stick to some of the Electronic Travel Aids, because some of them are either very bulky, expensive and many cannot provide effective detection of holes on the walking path. In [11] an obstacle detection system with multiple (infrared and ultrasonic) sensors was considered. The sensors were attached to the front part of the shoe, and the sensed signal is sent to a receiving part where it is processed to generate the required alert to the user. Its design, however, does not detect drop-offs. A good review of most of the existing ETA is provided in [12]. A recent design in [13] uses generated light patterns pointed at the field of view of a camera. The captured images of the matrix of the points is then analysed. The deformation of the pattern of light follows the object over which it is projected. Deep neural network consisting of convolutional neural network and Long short term memory units is then applied for classification of the object. The device developed is light and evaluated to have high accuracy, however it is not hands-free.

A method presented in [14] was a prototype developed on a walking stick for detecting obstacles and water puddles in the direction of the user. Another walking stick using an ultrasonic sensor and a moisture sensor was proposed in [15]. The ultrasonic sensor was attached to the stick at an angle of 40 degrees to detect holes or stairs. Though this can detect obstacles and stairs it cannot detect potholes because they have weak reflected waves. In [16], effort was made to address pothole detection. To achieve this, four ultrasonic sensors were arranged on a walking stick. Three of the sensors were arranged to sense obstacle on the left, right and front directions and the last sensor was placed below the walking stick to detect potholes. A predetermined distance was set between the sensor and the ground level. So a longer distance was expected when the stick passes over a hole and this increase in distance was used to detect the presence of a pothole. The problem with this approach is that this distance may vary and thereby give a false alarm while walking depending on the placement angle of the stick.

Most recent approaches employ the technique of computer vision or image processing in detecting objects in the path of the user. An approach considered in [17, 18] made use of the RGB-D camera and an ultrasonic sensor. Images of most commonly encountered standing obstacles were collected and stored in a database. The detection is then done by comparing what is captured by the camera with those in the database. The distance of the obstacle was measured by the ultrasonic sensor. The system was able to detect both obstacles and potholes and communicate via an audio message to the user. One problem with this is that objects not in the database may go undetected if encountered while walking. Another issue identified with the method was that it failed to classify the objects detected, so the user could not determine if what is detected is an obstacle or a pothole. Other contributions that made use of both vision and depth information for ETA and navigation sysem designs have been reported [19, 20, 21]. Another ETA that integrates both sensor-based and image processing techniques with multiple object detection was presented in [22]. The design is reported to have 100% accuracy in detection of obstacles, however, the design did not consider the detection holes and drop-offs.

It has been observed, in [23], that the visually impaired usually have their confidence increased when they have a better perception of their environment. Therefore, in addition to creating awareness of the presence of objects, it is necessary that people with visual impairments have an idea of the kind of objects (i.e. either obstacles or potholes) before them. It is also important to ensure that electronic mobility aids (EMAs) affordable and hands-free. These thus formed the major focus of this study.

2. The instrument design

The proposed system, as shown in Fig. 1, is based on the multi-sensor fusion approach to detection. The system is composed of three sub-units, namely: the sensing unit, the processing unit and the alerting unit. The sensing unit consists of three transducers: a camera, an ultrasonic module and a laser module. Upon these transducers are developed three techniques in the processing unit for the detection of depression/hole or obstacle. The first technique uses the ultrasonic module to detect obstacles. The second technique is designed around the laser module and the camera. It uses the methods of image processing to accomplishes the detection process. The third technique is solely based on the images from the camera and methods of image processing to detect obstacles or depression. While each of the sensing techniques can detect obstacles, only the first method, that uses the ultrasonic module cannot effectively detect depression or hole in the ground.

Figure 1.

Block diagram of the system.

2.1 The sensing unit

The sensing unit has all the three transducers purposely attached to a waist belt. The camera is positioned on the belt at an angle to the vertical to generate a red line on the ground at a distance of 1 m away from the feet of the user. Though the camera has a viewing range greater than 1.73 m and a viewing width of 2 m (i.e. 1 m each sideways), a shorter distance of 1 m and a width of 1.2 m is purposely used to ensure clear capture of the projected laser line. This is because of the normal adverse effect of the ambient illumination on the brightness and sharpness of the red line generated by laser. The ultrasonic sensor is also tilted to the vertical at the same angle with the camera.

2.2 The processing unit

The processing unit consists of a Raspberry Pi 3B which is connected to the three transducers of the sensing unit. It is also connected to the alerting unit. The three transducers connected to the Raspberry Pi are: USB camera, ultra-sonic sensor, earpiece. The USB camera which served as the image capturing device was connected directly to one of the two USB ports of the Raspberry Pi. The earpiece, serving as the user interface that relays distance information to the user, was connected to the audio/composite output jack of the Raspberry Pi. The Raspberry Pi and Laser device are both powered from a portable power bank. The other USB port of the Raspberry Pi was used to supply 5V from the power bank to it using a USB cable.

2.2.1 Technique 1: Using the ultrasonic module

An ultrasonic sensor is a transceiver module. The output pin (TRIG) of the ultrasonic sensor was connected to the GPIO 23 (Pin 16) of the Raspberry Pi while the input pin (ECHO) that reads reflected signal was connected to GPIO 24 (Pin 18) of the Raspberry Pi. The input voltage (i.e. 5 V) from ECHO pin of the ultrasonic sensor was limited to a safe operating voltage of 3.3 V using a voltage divider before connection to Pin 18 of the Raspberry Pi. It detects obstacles through the emission and reception of reflected short pulses from obstacles. It emits a short burst of 40 kHz which travels through air, at normal temperature, at a speed of 340 m/s. The ultrasonic sensor detects an obstacle on the floor at various distances away from the ultrasonic sensor. Thus, the ultrasonic sensor logically detects the presence of obstacle and also gives the estimate of the distance of the obstacle.

The actual distance was measured using a steel tape rule and was compared with the calculated distance, $x$ , from $L$ , the distance from the ultrasonic sensor to the obstacle, and $h$ , the height of the ultrasonic sensor from the floor, using the relation $x^{2}=L^{2-}$ h ${}^{2}$ . A maximum deviation of about 0.8 cm was recorded as shown in Table 1.

Table 1
Comparison of actual distance with measured distance

Actual distance (cm)	Measured distance (cm)	Deviation (cm)
7	6.99	0.01
9	8.90	0.10
20	20.11	0.11
30	29.84	0.16
60	59.63	0.37
120	119.98	0.02
140	140.01	0.01
180	179.51	0.49
200	199.25	0.75

2.2.2 Technique 2: Using the laser and the camera

The second technique detects standing obstacles and holes by using a source of laser light and a camera. The laser source is used to generate a reference ray of light, a straight horizontal red line at a certain distance away from the blind. From the principle of refraction of light, the red line tends to bend around obstacle or depression. Therefore, the red line is deformed on coming in contact with either depressed path (e.g. holes), as in Fig. 2), and standing obstacles as in Fig. 2b but remains not deformed on a free path as shown in Fig. 2a. The pattern of deformation is distinct for obstacles and holes as shown in Fig. 2. A camera is continuously focused to capture the scene in front of the blind. With a developed algorithm, the captured image is stored and duplicated into two images using a Raspberry Pi 3B as a processor.

Table 2
Decision table

Sensor/technique	Output status	Output status	Output status
Edge detection	0	1	1
Ultrasonic	0	0	1
With laser line	0	$-$ 1	1
Final decision	Path free	Hole detected	Obstacle detected

Figure 2.

The pattern of the red laser line (a) without deformation (b) deformation due to obstacle (c) deformation due to depression or drop-off.

2.2.3 Technique 3: Using only the camera

One of the images is extracted for the red line and the pattern is then used to detect and to differentiate obstacles from holes. In the duplicate image, the red line is suppressed to remove its edges leaving the remaining contents of the image. The remaining contents are analysed using edge detection to detect the presence of holes and obstacles. This is necessary to complement the laser mode for detection since laser light can be affected under a condition of extreme scene illumination. With the edge detection, unlike the laser deformation detection, the presence of obstacles is not differentiated from the presence of holes since both use edges as means of detection. However, the output of the edge detection compared with the output of the ultrasonic sensor is used to differentiate the two objects (i.e. holes from obstacles). The analysed outputs of the three detection mechanisms are combined to alert the blind via an audio cue (beeps). This is further explained in the next section. In addition, the ultrasonic sensor is also used in this study to measure the distance of the blind from a standing obstacle. Distance information is transmitted through patterns beeps that vary incrementally with changing barrier distance from the user.

2.2.4 Detection decision

A decision on which object is detected is determined based on the output of edge detection, the pattern of laser line and the output of the ultrasonic sensor. In this study, there are two outputs for edge detection. The first output is when the path is not free, i. e. an object either a pothole or an obstacle on the walk path is detected, and the second output for when the walk path is detected as free. The two state outputs are represented with binary digits 1 and 0 respectively. Since edges are used to detect both obstacles and potholes, a binary digit 1 is used to represent both objects. Unlike the detection by edges, laser line detection was used to give three distinct outputs. The first is when there is no deformation on the line (i.e. when the road is detected as free); the second is when an obstacle is detected and the third for when a pothole is detected. The state outputs are denoted by digits 0, 1 and $-$ 1 respectively. The pattern of refraction determines and differentiates the presence of obstacle path from a depressed path (e.g. a pothole path). The ultrasonic detection has only two state outputs. The last one is for when an obstacle is detected and the other output is given when a path is detected as free. The two states are denoted by 1 and 0 respectively. The decision table is as shown in Table 2.

2.3 The alerting unit

This unit basically is composed of a function that runs on the Raspberry Pi and an earpiece. The function generates beeping signal at different intervals based on the object detected from the result of combined detection decision and the distance of the detected object. The Raspberry Pi 3B has two audio output ports, namely, the HDMI socket and the headphone socket. Two beep frequencies were used, a lower frequency for an obstacle and a higher frequency for a hole detection. However, the number of beeps per time increases as the closeness to the object increases.

3. System evaluation

The complete system before it was packed in the waist pocket is as shown Fig. 3. A controlled trial experiment was carried out to objectively assess the improvement of the proposed electronic mobility aid through these parameters: object presence awareness (holes and obstacles), detection range and collision/falling rate. The trial was done with 20 (11 Male and 9 Female) volunteers blind users of various demographic data and with varied guide cane usage experience from the Federal Society of the Blinds, Lagos, Nigeria. The inclusion criteria were for the blind individuals within the age range of between 15 to 50 years old, and without other comorbidities that may impair mobility such as mental disorder, orthopaedics impairments, hearing impairments, etc. A randomized four artificial barrier courses were created, of which, the users negotiated two courses (one for obstacles and the other for holes) with the proposed electronic mobility aid and the other two courses with the guide cane following the same arrangement with the proposed electronic mobility aid.

Figure 3.

The Complete system.

Figure 4.

Participants during the trial sessions. (a) Detecting hole/ drop-off with the EMA; (b) Detecting obstacle with guide cane; (c) Detecting obstacles with the EMA.

3.1 Experiment terrain

A long tiled veranda of 20 m by 4 m with 4 laid out obstacles such as chairs, Tables, Plastic Drum, etc. was used for the experiment. To avoid map learning effect by the users, the positions of the obstacles were randomize thereby forming two barrier courses (i.e. obstacles). The tiled veranda enabled the accurate measurement of the detection range for the two devices. The other two barrier courses were the artificial holes of 0.61 m by 0.61 m by 0.30 m each created on a plain playing ground. The two pothole locations were randomized to prevent map learning effect by the users. The artificial hole was in essence to assess improvement in the holes detection capability of the proposed device. A boundary line of 0.4m to the respective holes was drawn beyond which no user was allowed to exceed for safety purposes.

Table 3
Experimental data of the 20 subject

S/N

Age

Sex

Minimum

obstacle

detection

distance with

the aid of a

cane for 20

participants

(m)

Minimum

obstacle

detection

distance with

the aid of an

EMA for 20

participants

(m)

Minimum

pothole

detection

distance

with the

aid of a

cane for 20

participants

(m)

Minimum

pothole

Detection

Distance

with the

aid of an

EMA for

participants

(m)

Number of

obstacles

collided

with in a

four

obstacle

course using

a cane

(i.e. out of

four obstacles)

Number of

obstacles

collided

with in

a four

obstacle

course using

an EMA

(i.e. out of four

obstacles)

Number

of hole

collided

with in a

single

course

using a

cane

Number

of hole

collided

with in a

single

course

using an

EMA

0.33

1.23

0.21

1.71

0.23

1.08

0.23

1.08

0.51

1.70

0.78

1.70

0.58

1.7

0.38

1.70

0.65

1.41

0.45

1.41

0.52

0.93

0.52

1.54

0.38

1.70

0.38

0.60

1.63

0.60

1.63

0.61

1.65

0.51

1.65

0.54

0.51

0.24

0.51

0.52

0.99

0.52

0.78

0.37

1.28

0.52

0.40

0.33

1.58

0.36

1.51

0.36

1.61

0.18

1.71

0.39

1.71

0.37

1.68

0.55

1.20

0.33

1.51

1.14

1.72

0.36

1.61

0.56

0.98

0.39

1.71

1.14

1.72

0.36

1.61

0.56

0.98

0.39

1.71

3.2 Training exercise

The volunteers who consented to participate in the experiment trials were trained on how to operate and on the usage of the proposed electronic mobility aid. No training was done on the usage of guide cane as they were already accustomed to that already. Two days were used altogether for both the training and the trial phase. The training exercise was repeated three times for each of the participants at different training sessions: Morning, Afternoon and Evening on the first day. The training guidelines involved the introduction of the electronic device and its mode of operation. Training on the mode of operation focused on how to fix the gadget to the waist, the earpiece to the ear, and how to identify and switch on the power button. Training on the beep patterns to be aware of barriers and the closeness especially for obstacles was also done. At the evening of the first day, a pre-trial phase was conducted as a learn and practice training exercise on a different site from the experiment site. To measure the distance of barrier detection, the participants were trained to pause movement once the first beeping sound is heard. The distance of detection can then be measured with a steel rule.

3.3 Trial session

This pre-trial phase was re-conducted in the early hours of the second day for remembrance, familiarization and for the users confidence with the device. Participants were asked to move forward, negotiating the barrier course from the initial position without given an explicit reference. In some occasions, participants did not maintain translational movement but in a trajectory form and hence, were given audio guidance by the trial supervisor towards the target direction. In this stage, the participants negotiated two barrier courses (obstacle and artificial hole) with their guide canes and the other two barrier courses (obstacles and artificial hole) with the proposed electronic device. Some of the participants during the trial sessions are as shown in Fig. 4. After the experiment, a survey was conducted through verbal interactions and questionnaire to sample the users views on the performance and the areas for amendments and possible improvements.

4. Results

In this study, the resulted numerical data were average for the 20 blind users who volunteered. Table 3 shows the data of the tests carried out with 20 blind volunteers. The p-values ( $p<$ 0:0001) were less than 0.0001 for all hypothesis and this means more than 99.99% level of confidence. A paired $t$ -test was adopted to compute these $p$ -values because the same blind users used both the proposed electronic mobility aid and the guide cane in the course.

4.1 Barrier detection distance

Table 4 compares the mean barrier detection distance for the guide cane and the proposed electronic mobility aid. The proposed electronic device has a significant average improvement on the barrier detection threshold distance with 1.36 $\pm$ 0.30 m over 0.49 $\pm$ 0.17 m with the guide cane for obstacle detection. There was a greater improvement in the pothole distance detection capability of the electronic mobility aid proposed over the guide cane as shown in Table 4. The limitation in the length of the guide cane led to the shorter mean range of barrier detection threshold. Also, from the mean obstacle and pothole distance detection of the guide cane, the results showed that there was a bit more level of difficulty in range detection of pothole (0.40 $\pm$ 0.13 m) than it was, compared to its obstacle detection distance (0.49 $\pm$ 0.17 m).

Table 4
Average barrier distance detection

Barriers	EMA (m)	Guide cane (m)	% Increase in detection
Obstacles	1.36 $\pm$ 0.30	0.496 $\pm$ 0.17	177.55 $\pm$ 0.47
Holes	1.33 $\pm$ 0.39	0.40 $\pm$ 0.13	232.50 $\pm$ 0.52

Table 5

Average collision/falling per detection

Barriers	EMA (%)	Guide cane (%)	% Decrease in detection
Obstacles	3.75 $\pm$ 2.15	20.00 $\pm$ 3.35	81.25 $\pm$ 5.50
Holes	10.00 $\pm$ 6.81	65.00 $\pm$ 10.67	84.62 $\pm$ 17.48

4.2 Barrier collision or falling rate

As shown in Table 5, the users with guide cane has an accuracy of detection of about 80% and therefore obstacle collision rate of 20.0 $\pm$ 3.35% on average as compared to 3.75 $\pm$ 1.99% with the proposed electronic mobility aid which has an accuracy of detection of about 96%. The collision rate with barriers such as obstacles significantly reduced by 83.25% with the developed electronic mobility aid as of most them were significantly detected at a safe distance greater than 1 m. Though, the participants were prevented from falling into the artificial hole for safety purpose, actual falling was recorded against the participants with high possibility of falling using the distance of first detection (or threshold distance) of 0.40 m (1.3ft) to the hole as a borderline for the yardstick. This implies that if the participant did not detect the hole before reaching of crossing this border-line, the participant is stopped from moving forward and this would be counted as actual falling into the hole. The results showed the guide cane at its worst having the average possibility of falling into the hole of 65.0 $\pm$ 10.67% compared with 10.00 $\pm$ 6.81% for the proposed electronic device. The proposed device showed a decline in the falling rate by 84.62%.

5. Observations and the feedback from the participants

It was discovered in the course of the exercise, the overall time taken on the barrier course was more on the Electronic Mobility Aid than was taken for the guide cane. It was so surprising to some of the users detecting barriers without physical contact. Hence, longer time was taken by some to familiarise with and to internalize the new sensory cue via beeps. In few scenarios, users did not recognize a change in beeps pattern which resulted into collision. However, as the experiment proceeded, users became more confident as they paid greater attention to the audio patterns. Interview conducted after the trials showed that all the 20 users believed the device is very effective at barrier detection.

About 61.23 % perceived the device to be better than guide cane while the rest believed the device is very good. No one said the device is less effective than guide cane. And about 75 % recommended the device to be reduced to a smaller size such that could be hardly seeing outside. All recommended that the device be insulated against water.

6. Conclusion

Novel barrier detection and a warning system for the blind to enhance mobility of the blind is presented in this work. From the results and the feedback from the users, it is believed that the EMA is more effective for both pothole and obstacle detection than the traditional guide cane as it offers longer range detection and collision or falling rate reduction without; and without physical contact.

Author contributions

CONCEPTION: Abimbola M. Jubril

PERFORMANCE OF WORK: Segun J. Samuel

INTERPRETATION OR ANALYSIS OF DATA: Segun J. Samuel

PREPARATION OF THE MANUSCRIPT: Abimbola M. Jubril and Segun J. Samuel

REVISION FOR IMPORTANT INTELLECTUAL CONTENT: Abimbola M. Jubril and Segun J. Samuel

SUPERVISION: Abimbola M. Jubril

Ethical considerations

The Institutional Review Board approval number for the research is IPHOAU/12/1585.

Footnotes

Acknowledgments

The authors wish to acknowledge the assistance of the authority and management of the Federal School of the Blind, Lagos, Nigeria.

Conflict of interest

The authors have no conflicts of interest to report.

References

WHO. Vision 2020: The Right to Sight: Global Initiative for the Elimination of Avoidable Blindness: Action Plan 2006-2011, Geneva, Switzerland. 2007.

Chaitali

Prasad

. Review paper on navigation system for visually, International Journal of Advanced Research in Computer and Communication Engineering. 2015; 4(1): 166-168.

Brabyn

. A Review of Mobility Aids and Means of Assessment in Electronic Spatial Sensing for the Blind. Dordrecht: Martinus Nijho Publishers, 1985.

Dowling

. Mobility Enhancement Using Simulated Artificial Human Vision, PhD thesis, Queensland University of Technology, 2007.

Oladayo

. A multidimensional walking aid for visually impaired using ultrasonic sensors network with voice guidance, International Journal of Intelligent Systems and Applications. 2014; 6(8): 53-59. doi: 10.5815/ijisa.2014.08.06.

Chaitali

Prasad

. Microcontroller based navigation system for visually impaired people, International Journal of Advanced Research in Computer and Communication Engineering. 2015; 4(8): 431-435.

Ramprabu

Gowthaman

. Smart cane for visually impaired people, International Journal of Computer Science and Information Technologies. 2013; 4(1): 24-28.

Dambhare

. Smart stick for blind: Obstacle detection, artificial vision and real-time assistance via gps, in: 2nd National Conference on Information and Communication Technology (NCICT). November 2011; (6): 31-33.

Singh

Paul

Mehra

Gupta

Sharma

Jain

, et al. Smart cane for the visually impaired: Design and controlled field testing of an affordable obstacle detection system, in: International Conference on Mobility and Transport for Elderly and Disabled Persons (TRANSED) June 1-4, 2010, Hong Kong, China. Hong Kong Society for Rehabilitation.

10.

Hossain

. Development of navigation aid for visually impaired persons, MSc. Thesis, International Islamic University Malaysia, 2010.

11.

Mustapha

Zayegh

Begg

. Multiple sensors based obstacle detection system, in: 4th International Conference on Intelligent and Advanced Systems, 12-14 June, 2012, Kuala Lumpur, Malaysia.

12.

Chanana

Paul

Balakrishnan

Rao

PVM

. Assistive technology solutions for aiding travel of pedestrians with visual impairment, Journal of Rehabilitation and Assistive Technologies Engineering. 2017; 4: 1-16. doi: 10.1177/2055668317725993.

13.

Cornacchia

Kakillioglu

Zheng

Velipasalar

. Deep learning-based obstacle detection and classification with portable uncalibrated patterned light, IEEE Sensors Journal. 2018; 18(20): 8416-8425. doi: 10.1109/JSEN.2018.2865306.

14.

Therib

. Smart blinding stick with holes,obstacles and ponds detector based on microcontroller. Journal of Babylon University/Engineering Science. 2017; 25(5): 1759-1768.

15.

Sahoo

Lin

Chang

. Design and implementation of a walking stick aid for visually challenged people. Sensors (Basel). 2019; 19(1): 1-17. Available from: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6339061. doi: 10.3390/s19010130.

16.

Hakim

Fadhil

. Navigation System for Visually Impaired People Based on RGB-D Camera and Ultrasonic Sens or. In: ICICT ’19: Proceedings of the International Conference on Information and Communication Technology, April, 2019. pp. 172-177. Available from: doi: 10.1145/3321289.3321303.

17.

Vanitha

Rajiv

Elangovan

Kumar

. A Smart walking stick for visually impaired using Raspberry pi. International Journal of Pure and Applied Mathematics. 2018; 119(16): 3485-3489.

18.

Sathya

Nithyaroopa

Betty

Santhoshni

Sabharinath

Ahana

. Smart walking stick for blind person, International Journal of Pure and Applied Mathematics. 2018; 118(20): 531-4536.

19.

Elmannai

Elleithy

. A highly accurate and reliable data fusion framework for guiding the visually impaired, IEEE Access. 2018; 6: 33029-33054.

20.

Abobeah

Hussein

Abdelwahab

Shoukry

. Wearable RGB Camera-based Navigation System for the Visually Impaired, in Proceedings of the 13th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications (VISIGRAPP 2018). 2018; 5: 555-562.

21.

Kanwal

Bostanci

Currie

Clark

. A navigation system for the visually impaired: a fusion of vision and depth sensor. Applied Bionics and Biomechanics. 2015; Article ID 479857, 15: 1-16.

22.

Aladrén

López-Nicolás

Puig

Guerrero

. Navigation assistance for the visually impaired using RGB-D sensor with range expansion. IEEE Systems Journal. 2016; 10(3): 922-932.

23.

Jeamwatthanachai

Wald

. Indoor navigation by blind people:Behaviors and challenges in unfamiliar spaces and buildings. British Journal of Visual Impairment. 2019; 37(2): 140-153. doi: 10.1177/0264619619833723.

A multisensor electronic traveling aid for the visually impaired

Abstract

BACKGROUND:

METHODS:

REDULTS:

Keywords

1. Introduction

2. The instrument design

2.2 The processing unit

2.2.1 Technique 1: Using the ultrasonic module

Table 1 Comparison of actual distance with measured distance

Table 2 Decision table

2.2.4 Detection decision

2.3 The alerting unit

3. System evaluation

Table 3 Experimental data of the 20 subject

3.3 Trial session

4. Results

4.1 Barrier detection distance

Table 4 Average barrier distance detection

5. Observations and the feedback from the participants

6. Conclusion

Author contributions

Ethical considerations

Footnotes

Acknowledgments

Conflict of interest

References

Table 1
Comparison of actual distance with measured distance

Table 2
Decision table

Table 3
Experimental data of the 20 subject

Table 4
Average barrier distance detection