II. Related Work Discussion
Several interesting research have been already done in this field; and both outdoor and indoor navigation systems for visually impaired and blind people been proposed. These systems are based on different approaches and technologies in order to provide functional variety and be useful to navigate in both known and unknown environment.
Proposed architectures include both so-‐called “wearable navigation systems” [19, 13, 21, 36] together with more mobile solutions when only one assisted portable device is required on the users side. Such portable devices might include a mobile unit [15], a smartphone [10, 2, 11], or even improved high-‐tech version of White cane [7, 39].
In [13] authors discuss necessity of building a navigation system that covers both indoor and outdoor case scenarios. They provide a description of “an integrated framework of a precise navigation system for visually impaired and blind people” [13]. Although some interesting ideas were proposed in this work, there is still a necessity for tests results, including performance and usability.
Studies [19] also propose outdoor and indoor wearable navigation system for visually impaired and blind people. For outdoor location positioning DGPS technique is used and special hardware equipment (Trimble PROXRS, 12 channel integrated GPS/Beacon/Satellite receiver with multi-‐path rejection technology) is introduced in the system in order to improve users outside location.
Koley et al. [21] introduces a navigation system for visually impaired persons, that combines usage of GPS, voice and ultrasonic sensor for obstacle detection. In the proposed system SiRFstarIII GR-‐301 GPS Receiver is used for outdoor location positioning. No GPS accuracy improvement techniques (RTK, DGPS, etc.) are used in order to improve outdoor location accuracy.
There are several systems designed particularly for traffic lights detection in order to help both color-‐blind drivers and visually impaired pedestrians [28, 29]. These systems are based on image recognition and video processing technologies.
As the purpose of this work is to provide an easy guidance assistance infrastructure that involves android-‐based mobile handheld device, the most relative works to this one also includes a mobile device and its equipped technologies like GPS and Networking modules.
One of the most relevant to this work research is [2]. The authors proposed cloud-‐based system architecture used together with an android-‐based mobile phone in order to detect traffic light status. This approach is based on video analysis techniques and requires a user to use GPS, Networking and Camera functionality on the mobile phone.
Proposed in [2] infrastructure includes two main components – a mobile device and a cloud server component. A mobile device computes its location using GPS satellites information and updates the cloud server with the current user location. It also serves as a video recording tool to capture traffic lights status. Cloud server component is responsible for processing the information received from the mobile device and sending back a response.
Studies [2] proposed architecture is shown in the Figure 1.
Figure 1 -‐ Studies [2] proposed system architecture
Although this work introduces interesting architecture, there are obviously several weaknesses that make it less efficient in comparison with so-‐called communication-‐based model2, introduced in this work.
In our opinion, video analysis or image processing based models would always introduce several important drawbacks, affecting the whole system usability and more importantly compromising the safety of a visually impaired/blind person.
One the biggest concern regarding the proposed architecture is an unhandled safety issue when a visually impaired or blind person is supposed to cross a road. Although, the authors discuss the importance of fast and reliable traffic light detection capability and mention real-‐time video frames processing, still the system lacks of actual real-‐time traffic light information status. Let us imagine the following case scenario: a traffic light is still in
“Green” state, but it is going to change its status to “Red” in 3 seconds. A mobile phone user captures traffic light (“Green”) and sends frame for analysis to the cloud. Obviously that even if the frame is analyzed immediately, and a person gets instruction to cross the road, there is still not enough time for them to comfortable and safely cross the road. This unavailability to predict traffic light status changes might introduce a great safety problem and compromise the whole system.
Another concern regarding proposed architecture is certain inconvenience for visually impaired/blind people to use the system. A requirement to record video along the path may lead to a big number of so-‐called “false records” when a user is not able not make camera adjustments manually. Particularly this may affect an ability to use camera in unknown environments, when exact traffic light locations and positions (including their height) are not obvious. Because of this uncertainty of traffic light locations, the architecture requires an introduction of additional assistance technology in order to easily adjust camera for successful video recording.
One more specific characteristic of the proposed in [2] architecture is related to outdoor location accuracy. As depicted in Figure 1, Skyhook Wireless positioning system is integrated in [2] architecture. Skyhook is a software-‐only location system that combines
2 The name of so-‐called communication-‐based model refers to the idea to make traffic lights status information available online
Wi-‐Fi positioning together with GPS and cell tower triangulation, its location's hybrid positioning accuracy is compared to both GPS and A-‐GPS in Figure 2.
Although, Skyhook provided services are flexible and fast, their location accuracy is never better than GPS. For some applications the fastest Time-‐to-‐Fix and constant availability is the most important characteristics in location services. These characteristics are also very important for blind navigation system, however better accuracy in location positioning detection is in the greatest demand.
Figure 2 -‐ Skyhook location accuracy
To sum up, all the previously mentioned issues introduce several obstacles in the proposed [2] architecture such as:
• Safety issue (real-‐time traffic light detection/prediction);
• Usage Inconvenience (necessity of using camera for a blind person);
• Location Accuracy
Therefore, the navigation system proposed in this Master thesis, is aimed to improve these weaknesses together with introducing new user case scenarios for outdoor navigation.