Sage Journals: Discover world-class research

Abstract

This study investigated the impact of secondary electronic travel aids (ETAs) on the mobility performance of individuals with profound visual impairment, specifically focusing on object detection and obstacle avoidance tasks. Seven experienced long-cane users were recruited for the study. The ETAs tested as secondary aids to the long cane included the MiniGuide, utilizing ultrasonic echo location, and the vOICe (Vision-to-Auditory Sensory Substitution Device), which translates visual information into auditory signals. Two software modes of the vOICe were examined: the standard contrast mode and a novel algorithm utilizing relative depth information. Notably, 30 min of familiarization was provided before the trials. Results indicated that while the use of the secondary ETAs led to increased awareness of the surrounding environment, it also resulted in longer processing times and more hesitations, compared with using the long cane alone. Participants demonstrated improved object detection and larger detection and safety ranges during obstacle avoidance. The study also introduced the concept of margin of stability (MoS) as a novel metric for evaluating gait stability while using ETAs. Both MiniGuide and vOICe users exhibited consistent and stable movements, suggesting that these additional aids did not negatively impact overall gait stability. The findings highlight the trade-offs between efficiency and effectiveness when using secondary mobility aids, with longer training periods potentially mitigating these issues. In addition, the study emphasizes the need for ongoing research to optimize the design and training programmes for new ETAs. Future studies may explore the impact of extended training periods on mobility performance. Overall, this research contributes to a deeper understanding of how ETAs can enhance the mobility of individuals with visual impairments while also highlighting some of the challenges associated with their use. Optimizing ETA design and training programmes is crucial for maximizing benefits for individuals with impaired vision.

Keywords

Electronic travel aid gait analysis low vision mobility outcome measures sensory substitution device

Mobility aids play an important role in enhancing the independence and safety of individuals with visual impairments, assisting their navigation through various environments (Deverell et al., 2015). Primary mobility aids are intended to be used independently and are equipped to detect obstacles and drop-offs in the travel path. The long cane is the most widely accepted primary aid and assists mobility by providing tactile feedback (Kim et al., 2009). While effective, the long cane has inherent limitations. For instance, users may struggle to detect overhanging obstacles, and approaching obstacles, particularly in complex or dynamic environments where accurate spatial awareness is crucial (Kim & Emerson, 2014). These limitations have led to the development of emerging aids, such as electronic travel aids (ETAs), which are designed to augment or complement primary aids to address their shortcomings (Kreilinger et al., 2018).

Among these emerging aids, MiniGuide (GDP Research, Adelaide, Australia) and vOICe (Vision-to-Auditory Sensory Substitution Device) (Meijer, 1992) represent promising technologies designed to assist people with severe visual impairment (F. E. Brown et al., 2019; Hill & Black, 2003). The MiniGuide is a compact ETA, introduced as a simple handheld device emitting ultrasonic signals to detect nearby obstacles. It provides vibratory feedback to the hand when it detects an obstacle, with the intensity of the vibration corresponding to the proximity of the obstacle. Previous survey results have indicated that MiniGuide users identified the device as useful in several life situations, including locating items such as rubbish bins, seats, and people, as well as finding gaps between objects like elevator doors or doorways (Hill & Black, 2003). The vOICe is a vision-to-auditory ETA that translates visual information into auditory signals, enabling individuals with visual impairments to perceive their surroundings through sound (Meijer, 1992). The vOICe converts camera-captured greyscale images into audio output, with the audio intensity corresponding to contrast levels and stereo panning conveying spatial information. Both groups, Auvray et al. (2007) and D. Brown et al. (2011) trained participants and observed that they could accurately locate and identify objects on a table using the vOICe. However, limited research has explored how these aids complement primary aids like the long cane, which they were designed to supplement, in enhancing mobility performance, particularly in tasks such as object detection and obstacle avoidance (Ward & Meijer, 2010). Object (i.e., landmark) recognition and obstacle avoidance are two distinct and fundamental tasks for individuals with visual impairments when navigating their surroundings (El-Taher et al., 2021). Object recognition involves identifying and categorizing objects in the environment, while obstacle avoidance demands detecting and safely navigating around obstacles to avoid collisions. Understanding the effectiveness of these additional aids and evaluating their impact when used in conjunction with the long cane is essential for optimizing their design to better meet the needs of users.

In addition to hardware advancements, developments in software technology have also led to the exploration of innovative adaptations of ETAs. For example, vOICe is an open-sourced technology that has inspired many adaptations – including the combined haptic and audio ‘Sound of Vision’ (SoV) system which reported a reduction in collisions and a higher awareness of obstacles for low-vision participants (Hoffmann et al., 2018). By further example, SoundSight integrates computer vision and mobile computing to highlight salient objects and deliver information on depth, colour, or thermal profile (Hamilton-Fletcher et al., 2022).

A significant portion of evaluation studies focused on subjective outcome measures, including patient reports and observer data (F. E. Brown et al., 2019; Hill & Black, 2003). With this continued advancement of ETAs, an important pre-requisite in this field is the establishment of evaluation methods to assess their behavioural effects quantitatively. The new evaluation tool should be able to explore the functional improvement of advanced aids, assess changes in user’s mobility performance, and generate objective comparisons in comparison to primary aids.

In this study, we aimed to evaluate the effectiveness of different mobility aids, including the MiniGuide, the vOICe with software adaptations, and the long cane, in assisting users with object detection and obstacle avoidance tasks. We employed gold-standard laboratory motion tracking techniques to objectively measure and analyse the mobility performance of participants in terms of efficiency, effectiveness, and gait stability. We captured detailed movement data, including speed, trajectory, and interactions with the environment, to provide valuable insights into strengths, weaknesses, and potential areas for improvement in mobility aid design and functionality.

Methods

Participants

Participants were recruited through the research databases of one of the authors, LNA. The inclusion criteria included: (1) permanent and profound vision loss (less than 6/60 OU); (2) prior experience in using the long cane for mobility assistance; and (3) absence of any gait or hearing-related issues. The study was approved by the Human Research Ethics Committee of the University of Melbourne (ID 2056772) and abided by the Declaration of Helsinki. Participants were reimbursed for their time and travel expenses with a $50 gift voucher. Prior to participation, participants were provided with detailed information about the study, including its objectives and procedures, through discussions. They were then able to provide informed consent.

Experimental setup

Two mobility aids were tested in addition to the long cane. The first was the vOICe, a vision-to-auditory ETA which transforms greyscale images into audio signals. The vOICe setup consisted of a head-mounted camera, utilizing a baseball cap for camera positioning (Intel RealSense D435i, Intel, USA) for video input, a backpack containing the processing unit (HP EliteBook x360 1030 G8, HP, USA), an in-ear headphone (Freedom 2, Jaybird, USA) for delivering auditory signals (Figure 1a). The second device under test was the MiniGuide, a handheld ETA, which uses ultrasonic echo location to detect objects (Figure 1b). The MiniGuide vibrates to indicate the distance to objects, with a faster vibration rate indicating a nearer object. For distances ranging from 0.5 to 2 m, the vibration frequency varies from 55 to 10 Hz. The maximum detection range for the MiniGuide was set to 2 m during testing. This distance was chosen to allow enough room for adjusting gait and to eliminate feedback from farther distances. In addition, participants were instructed to provide and use their own long cane in their preferred hand. In trials with the MiniGuide, the device was held in the opposite hand (Figure 1c). All participants were blindfolded during the experiment to achieve a baseline vision of light perception or worse.

Figure 1.

Experimental setup. (a) The vOICe setup consisted of a cap-mounted camera, a laptop worn in a light backpack, and in-ear headphones. (b) The MiniGuide and a bluetooth button registering the timestamp. (c) The experimental setup for this study (without the retroreflective markers).

The depth-sensing camera we utilized for the vOICe system (RealSense D435i, Intel, Santa Clara, CA, USA) not only captured standard RGB images but also depth images, where each pixel value represents the distance in space of the scene location projecting to that point (in millimetres). An image processing algorithm was applied to captured depth images, translating each image in real-time to a 72×72 display grid, which was then converted into audio signals utilizing the vOICe’s encoding algorithm. Two modes of display were trialled during the vOICe system testing (Figure 2). The first, referred to as ‘binary mode’, provided feedback at maximum intensity (i.e., maximum comfort level) for any point in the display grid belonging to an object within a 2-m range (as determined from the corresponding grid position value). The second mode of display trialled was ‘depth mode’, where feedback for points on objects within a 2-m range indicated the proximity of the object. Feedback gradually increased from minimal detection intensity (i.e., threshold intensity) for objects exactly 2 m away, to maximum intensity as the user approached the object. In both modes of display, feedback was set to 0 (i.e., no feedback) for any display grid position corresponding to surfaces more than 2 m away, thereby removing all background surfaces in the final audio-rendered display.

Figure 2.

Two image processing modes of the vOICe trialled in this study. In the binary mode, objects provide audio feedback at maximum intensity within a 2-m range. In the depth mode, audio feedback starts at minimal intensity at a distance of 2 m, gradually increasing as the user approaches.

Participants received 30 min of training before the trial. The training consisted of familiarization with the devices for 10 min, which included an introduction to the MiniGuide and vOICe, their components, and basic operation methods. This was followed by a 5-min session to calibrate the audio output of the vOICe, determining the maximum comfort level of sound volume for each participant (‘maximum comfort level’). Finally, a 15-min depth discrimination training ensured that participants could use both the MiniGuide and vOICe in depth mode to detect and differentiate objects at various distances. A minimal detection range was also established to ensure participants could detect objects within a designated 2-m range and have distinct perceptions at three different ranges: 0.5, 1, and 1.5 m for both vOICe and MiniGuide devices. Input camera angles for vOICe were calibrated to ensure consistent views of the testing field.

Kinematic data were collected using a 12-camera motion tracking system (Vicon Vero v2.2, Vicon Motion Systems Ltd, Oxford, UK) at 120 Hz. A full-body marker set, consisting of 40 14-mm retroreflective markers, was placed on the participants’ body based on the University of Western Australia (UWA) marker set (Supplemental Appendix 1) (Besier et al., 2003). Four retroreflective markers were also placed on the participants’ cane and on each obstacle.

Data collection

The experiment consisted of two tasks: an object detection task and an obstacle avoidance task. In the object detection task, participants were presented with an object placed at different distances and orientations. Participants were required to search and detect the object using the device under test, as well as their long cane. For the obstacle avoidance task, participants were instructed to navigate through a course containing obstacles of varying heights and widths. Participants were first familiarized with the objects. A verbal description of the items was provided, and participants were allowed to touch the objects. Participants were instructed to wear headphones to eliminate any audio cues. A starting cue, indicated by a beep through participants’ headphones, initiated each trial, which corresponded with the devices being turned on. There was a pre-defined pseudorandom table specifying the order of conditions used in the study.

Task 1: object detection

The object, a medium-sized cylinder (⌀ 33 cm, H 107 cm), was randomly positioned in front of the participant for each trial. Participants were instructed to complete the task using one of the four different assistance profiles: (1) only with the long cane; (2) long cane combined with the MiniGuide; 93) long cane combined with vOICe in binary mode; and (4) long cane combined with vOICe in depth mode. The position of the object and the mobility aids were randomized according to a predetermined table before the session. Each participant completed a total of 24 trials (4 assistance profiles × 6 positions). Participants were instructed to explore the area in front of them and locate the object using different aid profiles. They were asked to press the Bluetooth button attached to their left wrist when they detected the object and then to walk towards and touch the object to mark the end of the trial. The button captured the timestamp of detection event. The exploration area covered a 60° arc in front of participants’ initial position and the target obstacle was randomly placed at either 1.5 or 3 m (Figure 3a).

Figure 3.

Trial setups. (a) An illustration of trial setup for the object detection task. There were six potential object positions: 1.5 and 3 m away, positioned straight ahead, 30° to the left, and 30° to the right. (b) An illustration of trial setup for the obstacle avoidance task, viewed from above. The testing field measures 3 m in width and 6 m in length. The obstacles are positioned 2 m apart from each other.

Task 2: obstacle avoidance

Three obstacles, (1) a large rectangular box (L 34 cm, W 35 cm, H 80 cm); (2) a small bin (L 25 cm, W 25 cm, H 27 cm); and (3) a medium-sized cylinder (⌀ 33 cm, H 107 cm), were placed on the ground for each trial. The course was 6 m long, with obstacles positioned at intervals of 2 m along the course and up to 1 m from the midline (Figure 3b). Since vOICe in binary mode (vOICe-B) does not assist in discerning the relative distance between the user and obstacles, it was not included in the obstacle avoidance task. For each trial, participants were instructed to complete the task using one of the three different assistance profiles: (1) only with the long cane; (2) long cane combined with the MiniGuide; and (3) long cane combined with vOICe in depth mode. Participants were instructed to walk through the course from the start line to the finish line as efficiently as possible while avoiding contact with the obstacles. They were asked to press the Bluetooth button on their wrist when they detected an obstacle. The position of obstacles and assistance profiles were determined by a randomization allocation table. Each participant completed a total of 18 trials (3 assistance profiles × 6 course configurations).

Data analysis

Three-dimensional marker trajectories were pre-processed within Vicon Nexus (Version 2.12, Vicon Motion Systems Ltd., Oxford, UK) and filtered using a six-order zero-lag Butterworth low-pass filter. An illustrative example of participants’ trajectory and object location is provided in Supplemental Appendix 2. Initial analysis was performed using custom in-house scripts within BodyBuilder software (Version 3.6.4, Vicon Motion Systems Ltd, Oxford, UK). Dynamic movement of each participant was characterized by the whole-body centre of mass (CoM_WB) and extrapolated centre of mass (XCoM) (Pai & Patton, 1997). As the participants were performing highly variable movements (i.e., reaching and leaning), we quantified these metrics using a segmental method (Della Croce et al., 2005), where the CoM_WB was defined as the weighted sum of each segment’s centre of mass from regression equations (Tisserand et al., 2016). Briefly, the hip joint centre was calculated using a regression equation derived from the markers on the pelvis (Harrington et al., 2007), while the shoulder joint centre was defined using the UWA offset method (Elliott et al., 2007). Ankle, knee, elbow, and wrist joint centres were defined as the midpoint between two markers positioned on bony landmarks either side of the joint (malleoli, condyles, etc.). Using these joint centres, nine body segments were created with scaled sex-specific body segment inertial parameters (Dumas et al., 2007; Yakowitz & Szidarovszky, 1989) (trunk and bilateral shanks, thighs, upper arms, and forearms), with the most distal segments (head, hand, and foot) merged with their proximal segments (Tisserand et al., 2016). MoS was defined as the distance between the XCoM and the anterior limit of the base of support (BoS) (i.e., the position of the toes of the front feet) at the point of heel strike in the anterior-posterior direction (Watson et al., 2021). Cane movement was quantified as an angular velocity relative to the unilateral shoulder joint centre, with the assumption that the arm was rigid during movement. Post analysis was executed using customized scripts via Python (Python, Version 3.9.12), graphical user interface Jupyter Notebook (Project Jupyter, Version 7.0.4), MATLAB (MathWorks, Version R2022b), and SPSS (IBM, Version 29.0.0).

In the object detection task, participants’ performance was evaluated from three domains, including efficiency, effectiveness, and stability. Six outcome measures were used to assess these domains. Detailed explanations of each outcome measure can be found in Table 1. Efficiency was determined by the contact time and the hesitation time, reflecting how quickly participants found the object. Effectiveness referred to the task success rate, the average cane swing angle, and the deviation, showing the accuracy and precision of their actions. Stability was assessed via the MoS, which indicated the region of error tolerance in participants’ movements. The latter has been proposed to quantify stability during dynamic tasks such as walking, a positive value of this metric reflects positive stability (Hof et al., 2005).

Table 1.

Assessing mobility performance: key outcome measures.

Domain	Task 1: object detection		Task 2: obstacle avoidance
Efficiency	Measure 1: Contact time (s)	Duration from the start of the trial, to when the participant made physical contact with the object.	Measure 1: Passing time (s)	Duration participants required to walk from the starting point to the endline (6 m).
Efficiency	Measure 2: Hesitation time (s)	Total time during which the participant’s velocity fell below a specific threshold (0.1 m/s).	Measure 2: Hesitation time (s)	Total time during which the participant’s velocity fell below a specific threshold (0.1 m/s).
Effectiveness	Measure 1: Success rate (%)	Percentage of trials that participants successfully detected and touched the object using the aids.	Measure 1: Cane angle (°)	Average swinging cane angle, relative to the holding hand and shoulder.
	Measure 2: Cane angle (°)	Average swinging cane angles, relative to the holding hand and shoulder.	Measure 2: Deviating distance (m)	Distance between participants and an obstacle when they began to deviate from their original path to avoid the obstacle upon detection.
	Measure 3: Deviation (m)	Averaged perpendicular distance between the COM trajectory and the ideal straight-line displacement between the start point and the object.	Measure 3: Clearing distance (m)	Minimum distance between participants and an obstacle while passing it.
Stability	Measure 1: Margin of stability (cm)	Distance between the extrapolated COM and the support foot.	Measure 1: Margin of stability (cm)	Distance between the extrapolated COM and the support foot.

COM = centre of mass. The term ‘participants’ refers to the COM point, while ‘obstacle’ refers to the centre point of the obstacle.

In the obstacle avoidance task, participants’ performance was evaluated from efficiency, effectiveness, and stability. Efficiency was indicated by the passing time and the hesitation time. Effectiveness was determined by the average cane angle, the deviating distance, and the clearing distance, showing participants’ ability to successfully avoid obstacles without deviating significantly from the intended path. Stability referred to the MoS, indicating the consistency of participants’ gait patterns during obstacle negotiation.

For each outcome measure, descriptive statistics such as mean, standard deviation, and range were calculated to summarize participants’ performance. Friedman test was conducted to determine if there was a significant effect of mobility aids on each outcome measure. In addition, the post hoc analysis with the Wilcoxon test was conducted to determine which mobility aids were different from others. Bonferroni correction was applied to correct for multiple comparison errors. For all statistical tests, an alpha level of 0.05 was chosen.

Results

Participants demographics

Seven participants with profound low vision were recruited, with most having baseline best-corrected visual acuity of light perception. The participants had an average age of 42.4 years, with 5 of them being female. All participants were familiar with the long cane. Notably, 57% (4/7) used a guide dog as their primary mobility aid, with two using sighted guide (i.e., a friend or family member guiding them) and one using the long cane. None of the participants had used ETAs as primary aids for mobility assistance. The basic demographic information of the participants is presented in Table 2.

Table 2.

Basic demographic information of participants.

Participant	Sex	Age (years)	Visual acuity	Habitual primary mobility aid
P1	F	45	Light perception	Guide dog
P2	F	35	Light perception	Guide dog
P3	F	46	Light perception	Guide dog
P4	M	45	Light perception	Sighted guide
P5	M	30	Right eye: logMAR 1.50 Left eye: logMAR 1.60	Long cane
P6	F	54	Light perception	Sighted guide
P7	F	42	Light perception	Guide dog

Object detection task

Here we present the outcome measures for object detection tasks at distances of 1.5 and 3 m using different mobility aids: the long cane, the long cane with MiniGuide, the long cane with vOICe in binary mode (vOICe-B), and the long cane with vOICe in depth mode (vOICe-D). Descriptive analysis results for all outcome measures are presented in Table 3.

Table 3.

Outcome measures for object detection task.

	Contact time (s)	Hesitation (s)	Task success rate (%)	Average cane angle (degree)	Deviation (m)	Margin of stability (cm)
Object at 1.5m
Cane	9.78 ± 5.15	2.42 ± 1.34	85.71	13.34 ± 4.80	0.18 ± 0.17	1.60 ± 0.92
MiniGuide	11.24 ± 5.58	4.17 ± 3.45	100	9.16 ± 2.98	0.08 ± 0.06	1.84 ± 1.02
vOICe-B	18.48 ± 16.84	8.45 ± 11.43	100	9.05 ± 3.54	0.12 ± 0.14	2.95 ± 2.84
vOICe-D	15.95 ± 9.24	6.76 ± 7.06	100	9.98 ± 3.35	0.12 ± 0.10	2.52 ± 1.48
Object at 3m
Cane	20.64 ± 10.68	3.37 ± 1.78	57.12	13.67 ± 4.71	0.48 ± 0.34	3.39 ± 1.64
MiniGuide	28.01 ± 14.15	7.72 ± 6.91	100	10.34 ± 3.17	0.25 ± 0.16	4.45 ± 2.40
vOICe-B	42.71 ± 35.14	16.74 ± 24.39	100	10.21 ± 2.83	0.40 ± 0.31	7.31 ± 6.23
vOICe-D	35.52 ± 19.88	12.68 ± 14.21	95.21	9.05 ± 2.32	0.33 ± 0.21	5.82 ± 3.74

‘vOICe-B’ refers to vOICe in binary mode, while vOICe-D refers to vOICe in depth mode. The data are reported as mean ± standard deviation.

Efficiency

Outcome measure 1: contact time

For detecting objects at 1.5 m, significant differences were observed in contact time when participants used the long cane alone compared with its use in combination with other devices (Figure 4). Participants took longer to locate and contact objects when using the vOICe in both binary (Z = –3.59, p < .001) and depth modes (Z = –3.62, p < .001), compared with using the long cane alone. In addition, participants required significantly more time to complete the task when using the vOICe compared with using the MiniGuide (Z = –2.72, p = .006). For detecting object at 3 m, compared with using the long cane alone, both additional ETAs – MiniGuide (Z = –2.46, p = .007) and vOICe in binary mode (Z = –2.88, p = .002) – prolonged the time to complete the task.

Figure 4.

Outcome measures of efficiency in object detection. Contact time refers to the duration from the start of the trial until the participant makes contact with the object. Hesitation time represents the total duration during which the participant’s velocity is below a threshold (0.1 m/s). Data were excluded from the analysis when participants failed to detect the object.

Outcome measure 2: hesitation time

In terms of identifying object at 1.5 m, all 3 additional ETAs – MiniGuide (Z = –3.01, p = .001), vOICe in binary mode (Z = –2.66, p = .003), and vOICe in depth mode (Z = –3.02, p = .001) – led to significantly longer overall hesitation time compared with using only the long cane (Figure 4). There was no significant difference observed among the additional aids. Similarly, for identifying object at 3 m, all additional aids, MiniGuide (Z = –2.88, p = .002), vOICe in binary mode (Z = –2.88, p = .002), and vOICe in depth mode (Z = –2.58, p = .005), resulted in significantly increased overall hesitation time compared with using only the long cane.

Effectiveness

Outcome measure 1: task success rate

In the search for objects at 1.5 m, participants demonstrated high performance, with success rates exceeding 80% in all cases. With the inclusion of the MiniGuide and vOICe in binary mode, success rates reached 100% for all participants, indicating their effectiveness in aiding participants’ object detection.

When attempting to locate objects at a further distance of 3 m, participants had a success rate of 58% when relying solely on the long cane. However, with the support of the MiniGuide and vOICe, participants achieved success rates of nearly 100%, with a slightly higher incidence of errors noted in the usage of the vOICe in depth mode (95.2%).

Outcome measure 2: average cane swing angle

For detecting object at 1.5 m, participants exhibited significantly smaller average cane angles when using the long cane in conjunction with the MiniGuide (Z = –2.79, p = .002), vOICe in binary mode (Z = –2.42, p = .007), and vOICe in depth mode (Z = –2.45, p = .007) compared with using the long cane alone (Figure 5). Similar reductions in average cane angle were also observed when detecting objects at 3 m across all additional aids, including MiniGuide (Z = –2.42, p = .007) and vOICe in depth mode (Z = –2.79, p = .002).

Figure 5.

Outcome measures of effectiveness in object detection. Cane angle represents the average swing angle of the long cane used by participants during the task. Deviation refers to the average displacement distance between the participants’ actual trajectory and the ideal path, which is a straight line between the starting position and the object.

Outcome measure 3: deviation

Regarding deviation while searching for objects in near distances, no significant differences were observed between the various aids in terms of veering from the ideal path (Figure 5). An increased deviation distance was noted when searching for distant objects with the long cane compared with the MiniGuide (Z = –2.79, p = .002) and with the vOICe in binary mode compared with the MiniGuide (Z = –2.80, p = .002).

Stability

Outcome measure 1: margin of stability (MoS)

Participants using the vOICe in both binary (Z = –2.80, p = .002) and depth modes (Z = –2.79, p = .002) demonstrated significantly greater MoS compared with using the long cane alone, when searching for objects at 1.5 m (Figure 6), showing that their body was more stable when using the secondary ETAs. Breaking this down further, both vOICe binary (Z = –2.79, p = .002) and depth modes (Z = –2.41, p = .007) exhibited significantly greater MoS, compared with the MiniGuide. Similar results were observed in the search for objects at 3 m, with participants using the vOICe in binary mode (Z = –2.79, p = .002) showing larger MoS, compared with using the long cane alone. Participants using the vOICe in binary mode had larger MoS compared with those using the MiniGuide (Z = –2.56, p = .005)

Figure 6.

Outcome measures of stability in object detection. Margin of stability is the average distance between the extrapolated centre of mass and the boundaries of the base of support in the direction of movement, indicating the safety buffer to maintain balance.

Obstacle avoidance task

Here we present the outcome measures for obstacle avoidance using different mobility aids: the long cane, the long cane with MiniGuide, and the long cane with vOICe in depth mode (vOICe-D). Descriptive analysis results for all outcome measures are presented in Table 4.

Table 4.

Outcome measures for obstacle avoidance task.

	Passing time (s)	Hesitation (s)	Average cane angle (°)	Deviating distance (m)	Clearing distance (m)	Margin of stability (cm)
Cane	22.33 ± 11.03	2.33 ± 1.73	10.73 ± 2.90	1.12 ± 0.18	0.91 ± 0.19	4.25 ± 1.99
MiniGuide	38.34 ± 31.82	9.92 ± 17.26	9.80 ± 2.97	1.58 ± 0.33	1.17 ± 0.34	7.32 ± 7.23
vOICe-D	41.40 ± 36.03	15.01 ± 22.91	9.06 ± 2.09	1.49 ± 0.32	1.02 ± 0.19	8.29 ± 8.37

Efficiency

Outcome measure 1: passing time

Participants required significantly more time to complete the obstacle avoidance task when using additional aids such as the MiniGuide (Z = –2.79, p = .002) and vOICe in depth mode (Z = –2.79, p = .002) compared with using the long cane alone (Figure 7).

Figure 7.

Outcome measures of efficiency in obstacle avoidance. Passing time refers to the duration from the start of the trial to when participants reach the finish line. Hesitation time represents the overall time during which participants’ velocity remains below a threshold (0.1 m/s).

Outcome measure 2: hesitation time

In comparison to using the long cane alone, participants exhibited significantly more hesitation periods when using the MiniGuide (Z = –2.56, p = .005) and vOICe (Z = –2.41, p = .007) (Figure 7).

Effectiveness

Outcome measure 1: average cane angle

There was no significant difference in long cane swing angle metrics for obstacle avoidance trials when using the different secondary mobility aids. This suggests that participants maintained a consistent swinging pattern with the cane during the obstacle avoidance task regardless of the additional aids being used (Figure 8).

Figure 8.

Outcome measures of effectiveness in obstacle avoidance. Cane angle refers to the average swing angle of the long cane during the obstacle avoidance trial. Deviating distance represents the distance between participants and the obstacle at the point where they start deviating from their original path. Clearing distance is the minimal distance between participants and the obstacle when they successfully pass the obstacle.

Outcome measure 2: deviating distance

Despite the decrease in efficiency, participants using the additional aids demonstrated a higher level of effectiveness in avoiding obstacles (Figure 8). When avoiding obstacles during the task, participants showed significantly longer deviating distances (i.e., detection range) with additional help from the MiniGuide (Z = –2.56, p = .005) and vOICe in depth mode (Z = –2.78, p = .002) compared with using the long cane alone.

Outcome measure 3: clearing distance

Similar results were observed in the clearing distance to maintain safety, with the MiniGuide (Z = –2.56, p = .005) and vOICe in depth mode (Z = –2.66, p = .004) resulting in larger clearing distances.

Stability

Outcome measure 1: margin of stability (MoS)

Participants using the MiniGuide (Z = –2.79, p = .002) and vOICe in depth mode (Z = 2.83, p = .002) demonstrated significantly increased MoS, compared with using the long cane alone (Figure 9).

Figure 9.

Outcome measure of stability in obstacle avoidance. The margin of stability represents the average distance between the extrapolated centre of mass and the base of support (i.e., feet) during an obstacle avoidance task.

Discussion

Our study has shown additional information from the MiniGuide and vOICe ETAs, used as secondary devices alongside the long cane, could significantly change user behaviour, even with brief familiarization. First, the use of additional aids decreased efficiency, as evidenced by longer hesitation times and overall completion times. However, these aids improved effectiveness in detecting objects, resulting in higher success rates, especially at greater distances, and reduced swinging cane angle. A smaller swinging cane angle suggests that participants were receiving more sensory information from the secondary device and effectively integrating and acting upon it. In addition, the aids enhanced the detection and safety range during obstacle avoidance tasks. Finally, participants using additional aids, particularly the head-mounted vOICe, exhibited a larger MoS, indicating a more substantial safety buffer and overall better stability. Of particular note, the participants did not receive any structured training and had not used these devices previously.

While the addition of secondary ETAs decreased efficiency, by causing longer hesitation time, they also improved effectiveness in detecting and avoiding obstacles. This will enhance participants’ overall awareness of their surrounding environment while walking, which could lead to improved safety. These findings agree with previous research indicating that ETAs require longer processing times (Min Htike et al., 2021; Vincent et al., 2014). Additional aids will introduce additional cognitive load or decision-making processes, leading to more hesitations. Future research could also examine the number of hesitations and investigate them individually, analysing the feedback the user received and the reasons for the hesitations. Users and low-vision care providers need to balance between efficiency and effectiveness when choosing mobility aids. However, it is worth noting that the brief training period in our study, approximately 30 min, may have limited the participants’ ability to fully adapt to the ETAs. Future studies could investigate whether extended training, followed by post-study interviews, could enhance users’ trust and confidence in using the device.

We have observed that participants significantly reduced their swinging cane angle in the object detection task when using additional aids. However, in the obstacle avoidance task, the additional aids did not significantly change participants’ cane usage patterns. This indicates that while additional aids can enhance object detection efficiency, they do not alter the cane usage patterns in situations requiring more dynamic exploration and obstacle avoidance. Participants continued to rely on their established cane techniques regardless of the supplementary aids in these situations. This could be related to the distinct learning curves associated with using the ETA signal for searching for objects versus avoiding them, which may require more time to adapt (Yan et al., 2018). In addition, ETAs used in the current study have a relatively short detection range (2 m) due to restricted room space. Future studies could explore whether an extended detection range could lead to more precise object detection and avoidance.

The findings also highlight the limitations of the long cane, particularly in scenarios requiring detection of objects beyond initial reach and in navigating obstacles with a wider range, which also strengthens the importance of additional assistive technology can contribute to greater confidence and independence in mobility tasks. The results provided a way to quantitatively describe the integration of the long cane (e.g., swinging angle and speed) and ETAs (e.g., head- or hand-mounted devices, measured by hand swinging and head movement angles) and the ratio between the usage of primary and secondary aids could serve as a valuable indicator of how participants integrate these devices during navigation tasks. With longer training, we anticipate an increase in the usage of additional aids, reflecting a more balanced utilization of both primary and secondary devices. Tracking changes in this ratio over time could provide insights into the effectiveness of such training programmes.

The safety of users during independent mobility is a significant concern (Bowman & Liu, 2017), with traditional assessments focusing on obstacle interaction rather than on the dynamic stability of user’s body while using the device. Our study introduced the MoS as a novel metric informed by robust estimation of whole-body centre of mass to evaluate overall gait stability while using aids. Both MiniGuide and vOICe exhibited consistent and stable movements while navigating through the course, suggesting that the additional aids did not negatively impact participants’ overall gait stability. Future research could explore the correlation between MoS levels and users’ perceived confidence and safety levels while using the device, validating the measurements. It is worth noting that the increased MoS may be linked to longer hesitation periods, suggesting a trade-off between safety and efficiency. This is especially true when using multiple devices, which can significantly increase cognitive load and elevate the risk of not responding promptly to sudden changes in their path. Another explanation could be attributed to the handheld versus head-mounted usage of the device, with increased hand movement observed when using the cane and the MiniGuide. In contrast, head-mounted devices like vOICe may offer a more stable centre of mass. Device manufacturers should consider not only the functional efficacy but also impact on gait stability and user confidence.

Our testing was conducted in a controlled environment, which may not fully represent real-world usage. Participants tended to have a small average swinging angle (10°–15°) since they were aware they were in a controlled environment with specific goals. In real-world applications, using a long cane alone may require a larger swinging angle to detect challenges such as changes in terrain or drop-offs. However, the emerging use of wearable sensors to track human movement (i.e., with inertial measurement units) may provide avenues of future research to evaluate similar metrics in real-world scenarios. In addition, we have a small sample (N = 7) and most of the participants had baseline vision of light perception only, meaning that we have not fully explored the outcomes of these devices with a broader range of vision loss. Exploring how individuals with varying levels of remaining vision use different devices could provide valuable insights in real-world usage. It is important to note that although all participants were familiar with the long cane, only one of them used it as their primary mobility aid. For the rest of the participants, the long cane was used on limited occasions (e.g., uneven terrain) where additional assistance was needed. This distinction could have implications for their assisted mobility performance and therefore warrants further investigation. Another limitation of this study is that it focuses solely on performance improvement without considering real-world usability factors such as battery life, bulkiness, and price tag. In practice, participants tend to prefer portability and convenience, whereas our prototype involves a backpack. These factors should be evaluated as they are important for end-users in real-world scenarios.

Despite slight increases observed in hesitation with the vOICe binary mode, no significant differences were noted between the two vOICe modes. The short detection range and brief training period may not fully present the capabilities of the depth mode. Future research with longer detection ranges and extended training periods for the depth mode could provide improved insights. In addition, motion tracking techniques have the potential to not only monitor users but also understand how they interact with devices. Exploring how arm swinging with handheld devices and head movement with head-mounted devices are influenced by primary aids such as the long cane could optimize device usage strategies and rehabilitation programmes.

In conclusion, our study demonstrated the significant impact of integrating additional information from ETAs, such as MiniGuide and vOICe, alongside the long cane for individuals with visual impairments. Despite a brief familiarization period, participants showed behavioural changes characterized by increased awareness of their surroundings and enhanced stability in movement, but decreased efficiency and more hesitation when using additional aids. Motion tracking offered promising methods to quantify the trade-offs between efficiency, effectiveness, and overall gait stability when integrating additional aids with the long cane. In addition, the effect of extended experience and training is an important area for future research to further enhance the trust, confidence, and overall performance of users with these devices. Overall, our findings contribute to the understanding and optimization of mobility aid technologies.

Supplemental Material

sj-docx-1-jvi-10.1177_02646196241285098 – Supplemental material for Functional performance comparison of long cane and secondary electronic travel aids for mobility enhancement

Supplemental material, sj-docx-1-jvi-10.1177_02646196241285098 for Functional performance comparison of long cane and secondary electronic travel aids for mobility enhancement by Rui Jin, Matthew A Petoe, Chris D McCarthy, Jaime R Serra, Scott Starkey, Jennifer McGinley and Lauren N Ayton in British Journal of Visual Impairment

Supplemental Material

sj-docx-2-jvi-10.1177_02646196241285098 – Supplemental material for Functional performance comparison of long cane and secondary electronic travel aids for mobility enhancement

Supplemental material, sj-docx-2-jvi-10.1177_02646196241285098 for Functional performance comparison of long cane and secondary electronic travel aids for mobility enhancement by Rui Jin, Matthew A Petoe, Chris D McCarthy, Jaime R Serra, Scott Starkey, Jennifer McGinley and Lauren N Ayton in British Journal of Visual Impairment

Footnotes

Acknowledgements

We would like to thank Mr. Praveen Krishna for his assistance during the data collection stage of this project.

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by a National Health and Medical Research Council Investigator grant to LNA (GNT#1195713; 2022–26), University of Melbourne Driving Research Momentum Fellowship to LNA (2019–23), 2021 Retina Australia grant to LNA, MAP, CDM, and JLM, and a 2020 Melbourne Disability Institute grant to LNA, JLM, MAP, CDM, and RJ.

ORCID iD

Rui Jin

Supplemental material

Supplemental material for this article is available online.

References

Auvray

Hanneton

O’Regan

J. K.

(2007). Learning to perceive with a visuo-auditory substitution system: Localisation and object recognition with ‘the vOICe’. Perception, 36(3), 416–430.

Besier

T. F.

Sturnieks

D. L.

Alderson

J. A.

Lloyd

D. G.

(2003). Repeatability of gait data using a functional hip joint centre and a mean helical knee axis. Journal of Biomechanics, 36(8), 1159–1168.

Bowman

E. L.

Liu

(2017). Individuals with severely impaired vision can learn useful orientation and mobility skills in virtual streets and can use them to improve real street safety. PLOS ONE, 12(4), Article e0176534.

Brown

Macpherson

Ward

(2011). Seeing with sound? Exploring different characteristics of a visual-to-auditory sensory substitution device. Perception, 40(9), 1120–1135.

Brown

F. E.

Sutton

Yuen

H. M.

Green

Van Dorn

Braun

Cree

A. J.

Russell

S. R.

Lotery

A. J.

(2019). A novel, wearable, electronic visual aid to assist those with reduced peripheral vision. PLOS ONE, 14(10), Article e0223755.

Della Croce

Leardini

Chiari

Cappozzo

. (2005). Human movement analysis using stereophotogrammetry: Part 4: Assessment of anatomical landmark misplacement and its effects on joint kinematics. Gait & Posture, 21(2), 226–237.

Deverell

Bentley

S. A.

Ayton

L. N.

Delany

Keeffe

J. E.

(2015). Effective mobility framework: A tool for designing comprehensive O&M outcomes research. Vision Rehabilitation International, 7(1), 74–86.

Dumas

Cheze

Verriest

J.-P.

(2007). Adjustments to McConville et al. and Young et al. body segment inertial parameters. Journal of Biomechanics, 40(3), 543–553.

Elliott

B. C.

Alderson

J. A.

Denver

E. R.

(2007). System and modelling errors in motion analysis: Implications for the measurement of the elbow angle in cricket bowling. Journal of Biomechanics, 40(12), 2679–2685.

10.

El-Taher

E.-Z.

Taha

Courtney

Mckeever

. (2021). A systematic review of urban navigation systems for visually impaired people. Sensors, 21(9), 3103.

11.

Hamilton-Fletcher

Alvarez

Obrist

Ward

(2022). SoundSight: A mobile sensory substitution device that sonifies colour, distance, and temperature. Journal on Multimodal User Interfaces, 16, 107–123.

12.

Harrington

Zavatsky

Lawson

Yuan

Theologis

(2007). Prediction of the hip joint centre in adults, children, and patients with cerebral palsy based on magnetic resonance imaging. Journal of Biomechanics, 40(3), 595–602.

13.

Hill

Black

(2003). The Miniguide: A new electronic travel device. Journal of Visual Impairment & Blindness, 97(10), 1–6.

14.

Hof

A. L.

Gazendam

Sinke

(2005). The condition for dynamic stability. Journal of Biomechanics, 38(1), 1–8.

15.

Hoffmann

Spagnol

Kristjánsson

Á.

Unnthorsson

(2018). Evaluation of an audio-haptic sensory substitution device for enhancing spatial awareness for the visually impaired. Optometry and Vision Science, 95(9), 757–765.

16.

Kim

D. S.

Emerson

R. W.

(2014). Effect of cane technique on obstacle detection with the long cane. Journal of Visual Impairment & Blindness, 108(4), 335–340.

17.

Kim

D. S.

Emerson

R. W.

Curtis

(2009). Drop-off detection with the long cane: Effects of different cane techniques on performance. Journal of Visual Impairment & Blindness, 103(9), 519–530.

18.

Kreilinger

Georgi

Pregartner

Ivastinovic

Pichler

Berghold

Velikay-Parel

(2018). Quantifying the impact on navigation performance in visually impaired: Auditory information loss versus information gain enabled through electronic travel aids. PLOS ONE, 13(4), Article e0196156.

19.

Meijer

P. B.

(1992). An experimental system for auditory image representations. IEEE Transactions on Biomedical Engineering, 39(2), 112–121.

20.

Min Htike

H. H.

Margrain

Lai

Y.-K.

Eslambolchilar

. (2021, May 8–13). Augmented reality glasses as an orientation and mobility aid for people with low vision: A feasibility study of experiences and requirements [Conference session]. 2021 CHI Conference on Human Factors in Computing Systems, New York, NY, United States.

21.

Pai

Y.-C.

Patton

(1997). Center of mass velocity-position predictions for balance control. Journal of Biomechanics, 30(4), 347–354.

22.

Tisserand

Robert

Dumas

Cheze

(2016). A simplified marker set to define the center of mass for stability analysis in dynamic situations. Gait & Posture, 48, 64–67.

23.

Vincent

Routhier

Martel

Mottard

M.-E.

Dumont

Côté

Cloutier

(2014). Field testing of two electronic mobility aid devices for persons who are deaf-blind. Disability and Rehabilitation: Assistive Technology, 9(5), 414–420.

24.

Ward

Meijer

(2010). Visual experiences in the blind induced by an auditory sensory substitution device. Consciousness and Cognition, 19(1), 492–500.

25.

Watson

Fino

P. C.

Thornton

Heracleous

Loureiro

Leong

J. J.

(2021). Use of the margin of stability to quantify stability in pathologic gait–a qualitative systematic review. BMC Musculoskeletal Disorders, 22, Article 597.

26.

Yakowitz

Szidarovszky

(1989). An introduction to numerical computations. Macmillan Publishing.

27.

Yan

Zhaoping

(2018). Bottom-up saliency and top-down learning in the primary visual cortex of monkeys. Proceedings of the National Academy of Sciences, 115(41), 10499–10504.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.02 MB

0.37 MB