COST-EFFECTIVE MICROCONTROLLED OPTOELECTRONIC PORTABLE COLOR IDENTIFIER FOR VISUAL IMPAIRED PERSONS J. M. S. Pena, 1 C. Va ´ zquez, 1 I. Pe ´ rez, 1 X. Quintana, 2 and J. M. Oto ´n 2 1 Grupo Displays y Aplicaciones Foto ´ nicas Dpt. Tecnologı ´a Electro ´ nica, Escuela Polite ´ cnica Superior Universidad Carlos III c/Butarque 15, E-28911 Legane ´ s, Madrid, Spain 2 Grupo CLIQ de Cristales Lı´quidos Dpt. Tecnologı ´a Foto ´ nica Escuela T.S. Ingenieros de Telecomunicacio ´n Universidad Polite ´ cnica Ciudad Universitaria s/n E-28040 Madrid, Spain Received 30 April 2002 ABSTRACT: A hand-held color-identifier for blind, color-blind and vision-impaired people has been developed. Color data are obtained by balancing the relative intensities of RGB primaries. After a training pe- riod, the system is able to identify any arbitrary number of colors. The output is a spoken message, chosen from a preset list. © 2002 Wiley Periodicals, Inc. Microwave Opt Technol Lett 35: 309 –310, 2002; Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/mop.10592 Key words: optoelectronic system; portable color identifier; spoken message; visually handicapped INTRODUCTION Impaired color vision is a relatively frequent pathology, arising from persons with impaired vision, color-blindness, and other alterations. An approximate, if not exact, idea of the color of surrounding objects may contribute to improve the life quality of these persons [1]. In this context, color recognition may be useful for dressing, assessing food quality, location of warnings, etc. Several color systems have been proposed for blind people. The system designed by Furuno [2] showed a high accuracy on the most tested objects (85%–90%) and it used a tungsten filament lamp as emitter and three light sensors as optical detectors. This system requires six amplifiers to measure the optical signal coming from the lamp (3) and the reflected by the object (3), each one of them associated to the output signal of a primary color sensor. This approach usually presents a higher power consumption (provided that it needs a high number of electronic and optical components), and therefore, a shorter working time as compared to the device described in this paper. On the other hand, the Furuno’s system is only able to distinguish twelve colors. The other reported systems [3, 4] show fairly good performances such as the discrimination of high number of colored surfaces, synthetic speech, a wide range of interface options, among others. However, in the EU at least, these prototypes become too much expensive ($300 –$1000) for the applications above mentioned [4]. A simple color identifier has been developed. The aim of its design has been to obtain an inexpensive, lightweight, portable, user-friendly system. Taking into account the visual limitations of their users, spoken output has been chosen. Unless some special coding is adopted, this output limits in practice the number of distinguishable colors to the number of existing words describing such colors. Obviously the limitation depends on the language, and implies a specific calibration for each case. Yet if these difficulties are assumed, a simple portable device may be prepared, being useful in most instances where color identification is required. SYSTEM DESCRIPTION The color identifier is based on sequential reading of colored light intensities originated by three high efficiency LEDs. The most relevant features of these LEDs are summarized in Table 1. Se- quential reading of LEDs avoids the need of spectral separation of output light. Detection is performed by a TSL-230 light-frequency converter. LEDs and detector are managed by a PIC16C84 micro- controller. Unlike other complex color sensors which are able to provide true color recognition [5, 6], the proposed prototype only evaluates proportions of the RGB values received by the sensor, but not intensity. The intensity mode is intended for applications where a high level of color discrimination is required. However, in our case, this is not strictly necessary. The device geometry (Fig. 1) has been designed so that no straight light from any LED may reach the detector. The usual working mode is scattering mode. The color identifier is placed on the surface whose color is to be sensed. The device features a 2,5 cm circular aperture where the optoelectronic elements (LEDs and light converter) are located (Fig. 2). A soft cloth around the aperture largely reduces the ambient light. Three bursts of every color are generated and their readings are averaged. The light background is simultaneously sampled as well. A self-calibration procedure is performed on every acquisition. RESULTS AND DISCUSSION Self calibration allows the background to be corrected before interpreting the results. As a consequence, the device can read the color of objects located some cm away from its reading aperture (2–5 cm depending on the object surface). The color coordinates of the emitters have been computed. The LEDs cover an ample area of the CIE color diagram (Fig. 3), allowing most color variations to be potentially detectable [7]. A prototype has been prepared and trained using a large number of color samples taken from standard color collections. The readings have been assigned to names of colors, and these names have been stored in a look-up table. Sixteen named colors have been used in the prototype: Black Dark gray Light gray White Red Crimson Orange Magenta Yellow Brown Light green Dark green Violet Indigo Blue Cyan Color discrimination of the prototype is higher than this. In- deed, up to 64 colors could be separated readily. However, the use TABLE 1 Optoelectronic Characteristics of LEDs Employed in the Device LED Model  (nm)  (nm) Intensity (mcd) Luminous Efficiency (lm/W) Power Dissipation (mW) red HLMP-C100 644 18 750 85 100 green HLMP-CM30 526 47 1750 520 120 blue HLMP-CB30 472 35 560 75 120 MICROWAVE AND OPTICAL TECHNOLOGY LETTERS / Vol. 35, No. 4, November 20 2002 309