 |
|
|
|
A Virtual White Cane for the Visually Impaired
Roberto Manduchi, Computer Engineering
Professor Roberto Manduchi's research group in UCSC's Department of Computer Engineering develops algorithms for computer analysis of images and other sensor information. One project involves design of a hand-held laser sensor and processor, which together enable the visually impaired to efficiently and conveniently sense their external evironment. |
| |
The layout for a laser triangulation system used for sensing 3 dimensional objects.
|
|
The time profile of range acquired as the device was pivoted in an upward motion in front of the scene shown in the bottom figure (the dashed red line indicates the trace of the laser). The sensor was at a height of about 50 cm from the floor.
|
| x |
The trace generated by the laser beam as the user pivots the device in an upward motion around a horizontal axis in front of two steps. |
The most common mobility device for the blind is still the traditional long cane. This economical, simple, and reliable tool allows the user to extend touch and "preview" the lower portion of the foreground as one walks. However, not all visually impaired individuals are able to use the long cane. Furthermore, a rigid cane is an "invasive" tool, and often ill-suited to social gatherings, public transportation, or congested areas where it may trip pedestrians.
To address such problems, Manduchi's engineering group has developed a prototype mobility tool that adds a high-tech twist to the "long cane paradigm." In this case, instead of using a physical cane, a user scans the scene with a laser-based range sensing device. As the user moves the hand-held laser about, an onboard processor analyzes and integrates spatial information into a "mental image" of the scene, which is delivered to the user by means of a tactile interface.
Creating a devise that the visually impaired can trust is more difficult than one might first assume. Measuring distances to surfaces in a scene can help a user make decisions about where to move, but it may not be enough to move safely through the environment. For example, important features, such as curbs, steps, or drop-offs often go undetected by static range measurements.
The new system being developed by Manduchi's research group is designed around a hand-held laser sensor, coupled with a camera, that is swung in a pivoting motion to generate a steady stream of measured range data, which is much more efficient in detecting complex obstacles.
Another problem with many new detection systems is the latency of tactile interfaces used in many new detector systems. Manduchi's system therefore not only registers complicated environmental features, but includes an on board processor (an Extended Kalman Filter tracker) with algorithms that rapidly and accurately translate raw range data in a fashion that can be quickly and easily recognized by the user. In the recently tested prototype, sound is generated by an accompanying computer, with pitching varying with distance to the obstacle. Future versions are also being designed with a tactile interface, using signals called haptic primitives (e.g. small electrical pulses felt by the user).
|
