Blindness is a sensory disability that affects the ability to perceive light and, in severe cases, shapes and colors. This impairment, combined with inadequate infrastructure, hinders the mobility of blinds, both indoors and outdoors. Most of the assistive technologies for visual impaireds detect obstacles at the ground level. However, suspended objects such as cantilevers or a tree branch remain undetectable. This paper presents an embedded system that overcomes it, allowing to detect obstacles of these characteristics. The proposed system uses two kinds of proximity sensors, infrared (IR) and ultrasonic (US), arranged by pairs, pointing in three directions: left, forward, right. The angle between the pair of sensors pointing to the front and a side is 15 degrees. That setup allows measure three distances simultaneously while a blind person navigates through an unknown environment. Moreover, as will be shown below, the combination of these modalities reduces the uncertain of the measurements while helps to filter out the noisy samples. This paper also presents a multimodal data fusion algorithm that combines US and IR measurements into a single distance value, following a late fusion strategy. The proposed system was mounted on a like hat structure to facilitate its manufacture and validation. It also includes a haptic bracelet that feedbacks the depth perception through the vibration intensity of three motors, each ones indicating the distance to the nearest obstacle in its line of sight (left, forward, and right). Finally, experimental results have shown the usefulness of the proposed approach and its effectiveness as an assistance device for the visually impaired.