No abstract

Wireless capsule endoscopy (WCE) has become a good therapeutic method for a period of time. It helps detect, exam and heal gastro-intestinal (GI) related diseases. In the Capsule endoscopy application, knowledge of capsule position inside human body is rather important because it enables doctors locate the tumor of bleeding inside GI track and prepare for further therapeutic operations. However, due to the harsh environment for in-body wireless channel, in-body localization remains difficult and erroneous. In this paper, an improved three dimensional maximum likelihood algorithm has been introduced based on received signal strength (RSS) localization technology. Human body mesh and GI track mesh are built as the environment of algorithm simulation. Algorithm performance has been evaluated by comparison with the Cramer-Row Lower Bound (CRLB) and ranging error of the algorithm varies from 25mm to 140mm. By analyzing the results, we conclude that the three dimensional maximum likelihood is heavily impacted by the distance between implant and base station and its performance can be further improved.

Performance evaluation of wireless access and localization is important for body sensor networks, as any defects in the design not only cause wastage of resources, but also threaten an individual's health and safety. The typical cyber methods, however, such as software simulation, often fail to accurately simulate the influence of hardware implementation. The traditional physical methods, however, such as field testing, are not capable of creating repeatable and controllable channel conditions. To combine cyber and physical factors as well as to address the issue, we present a cyber physical test-bed for environment virtualization to facilitate the performance evaluation of wireless access and localization in body sensor networks. This test-bed creates a virtualized environment by emulating the wireless channel in a cybernetic way using a real time channel emulator. The original devices or systems under testing can be physically connected to a channel emulator to evaluate the performance in the virtualization environment. Furthermore, the cyber physical test-bed supports various scenarios from in-body data transmission to time of arrival based indoor localization. To validate the cyber physical approach, emulated outputs are compared with the empirical data obtained from actual measurements. To overcome the bandwidth limitation of traditional digital channel emulators, we have designed an analog channel emulator for UWB technologies. The preliminary verification of this analog emulator is introduced at the end of this paper.

Human activity recognition (HAR) problems have traditionally been solved by using engineered features obtained by heuristic methods. These methods ignore the time information of the streaming sensor data and cannot achieve sequential human activity recognition. With the use of traditional statistical learning methods, results could easily plunge into the local minimum other than the global optimal and also face the problem of low efficiency. Therefore, we propose a hybrid deep framework based on convolution operations, LSTM recurrent units, and ELM classifier; the advantages are as follows: (1) does not require expert knowledge in extracting features; (2) models temporal dynamics of features; and (3) is more suitable to classify the extracted features and shortens the runtime. All of these unique advantages make it superior to other HAR algorithms. We evaluate our framework on OPPORTUNITY dataset which has been used in OPPORTUNITY challenge. Results show that our proposed method outperforms deep nonrecurrent networks by 6%, outperforming the previous reported best result by 8%. When compared with neural network using BP algorithm, testing time reduced by 38%.