Tsung-Nan Lin

Indoor Location System Based on Discriminant-Adaptive Neural Network in IEEE 802.11 Environments

This brief paper presents a novel localization algorithm, named discriminant-adaptive neural network (DANN), which takes the received signal strength (RSS) from the access points (APs) as inputs to infer the client position in the wireless local area network (LAN) environment. We extract the useful information into discriminative components (DCs) for network learning. The nonlinear relationship between RSS and the position is then accurately constructed by incrementally inserting the DCs and recursively updating the weightings in the network until no further improvement is required. Our localization system is developed in a real-world wireless LAN WLAN environment, where the realistic RSS measurement is collected. We implement the traditional approaches on the same test bed, including weighted k -nearest neighbor (WKNN), maximum likelihood (ML), and multilayer perceptron (MLP), and compare the results. The experimental results indicate that the proposed algorithm is much higher in accuracy compared with other examined techniques. The improvement can be attributed to that only the useful information is efficiently extracted for positioning while the redundant information is regarded as noise and discarded. Finally, the analysis shows that our network intelligently accomplishes learning while the inserted DCs provide sufficient information.

Performance comparison of indoor positioning techniques based on location fingerprinting in wireless networks

Appropriate and correct indoor positioning in wireless networks could provide interesting services and applications in many domains. There are time of arrival (TOA), time difference of arrival (TDOA), angle of arrival (AOA), and location fingerprinting schemes that can he used for positioning. We focus on location fingerprinting in this paper since it is more applicable to complex indoor environments than other schemes. Location fingerprinting uses received signal strength to estimate locations of mobile nodes or users. Probabilistic method, k-nearest-neighbor, and neural networks are previously proposed positioning techniques based on location fingerprinting. However, most of these previous works only concentrate on accuracy, which means the average distance error. Actually, it is not enough to measure the performance of a positioning technique by the accuracy only. A comprehensive performance comparison is also critical and helpful in order to choose the most fitting algorithm in real environments. In this paper, we compare comprehensively various performance metrics including accuracy, precision, complexity, robustness, and scalability. Through our analysis and experiment results, k-nearest-neighbor reports the best overall performance for the indoor positioning purpose.

A Novel Algorithm for Multipath Fingerprinting in Indoor WLAN Environments

Positioning in indoor wireless environments is growing rapidly in importance and gains commercial interests in context-awareness applications. The essential challenge in localization is the severe fluctuation of receive signal strength (RSS) for the mobile client even at a fixed location. This work explores the major noisy source resulted from the multipath in an indoor wireless environment and presents an advanced positioning architecture to reduce the disturbance. Our contribution is to propose a novel approach to extract the robust signal feature from measured RSS which is provided by IEEE 802.11 MAC software so that the multipath effect can be mitigated efficiently. The dynamic multipath behavior, which can be modeled by a convolution operation in the time domain, can be transformed into an additive random variable in the logarithmic spectrum domain. That is, the convolution process becomes a linear and separable operation in the logarithmic spectrum domain and then can be effectively removed. To our best knowledge, this work is the first to enhance the robustness to a multipath fading condition, which is common in the environments of an indoor wireless LAN (WLAN) location fingerprinting system. Our approach is conceptually simple and easy to be implemented for practical applications. Neither a new hardware nor an extra sensor network installation is required. Both analytical simulation and experiments in a real WLAN environment demonstrate the usefulness of our approach to significant performance improvements. The numerical results show that the mean and the standard deviation of estimated error are reduced by 42% and 29%, respectively, as compared to the traditional maximum likelihood based approach. Moreover, the experimental results also show that fewer training samples are required to build the positioning models. This result can be attributed to that the location related information is effectively extracted by our algorithm.

Location Fingerprinting In A Decorrelated Space

We present a novel approach to the problem of the indoor localization in wireless environments. The main contribution of this paper is fourfold: 1) we show that by projecting the measured signal into a decorrelated signal space, the positioning accuracy is improved, since the cross correlation between each AP is reduced, 2) we demonstrate that this novel approach achieves a more efficient information compaction and provides a better scheme to reduce online computation (the drawback of AP selection techniques is overcome, since we reduce the dimensionality by combing features, and each component in the decorrelated space is the linear combination of all APs; therefore, a more efficient mechanism is provided to utilize information of all APs while reducing the computational complexity), 3) experimental results show that the size of training samples can be greatly reduced in the decorrelated space; that is, fewer human efforts are required for developing the system, and 4) we carry out comparisons between RSS and three classical decorrelated spaces, including Discrete Cosine Transform (DCT), Principal Component Analysis (PCA), and Independent Component Analysis (ICA) in this paper. Two AP selection criteria proposed in the literature, MaxMean and InfoGain are also compared. Testing on a realistic WLAN environment, we find that PCA achieves the best performance on the location fingerprinting task.

Principal Component Localization in Indoor WLAN Environments

This paper presents a novel approach to building a WLAN-based location fingerprinting system. Our algorithm intelligently transforms received signal strength (RSS) into principal components (PCs) such that the information of all access points (APs) is more efficiently utilized. Instead of selecting APs, the proposed technique replaces the elements with a subset of PCs to simultaneously improve the accuracy and reduce the online computation. Our experiments are conducted in a realistic WLAN environment. The results show that the mean error is reduced by 33.75 percent, and the complexity by 40 percent, as compared to the existing methods. Moreover, several benefits of our algorithm are demonstrated, such as requiring fewer training samples and enhancing the robustness to RSS anomalies.

IL28B SNP rs12979860 is a critical predictor for on-treatment and sustained virologic response in patients with hepatitis C virus genotype-1 infection.

Background Single nucleotide polymorphisms (SNPs) of interleukin-28B (IL28B) have received considerable interest for their association with sustained virological response (SVR) when treating patients of genotype-1 hepatitis C virus (GT1-HCV) chronic infection with pegylated interferon and ribavirin (PegIFN/RBV). This study was to investigate the predictive power of IL28B SNPs for on-treatment responses and SVR in treatment-naïve patients with GT1-HCV chronic infection. Methodology/Principal Findings We analyzed ten SNPs of IL28B in 191 treatment-naïve patients with GT1-HCV chronic infection who received PegIFN/RBV. In these patients, rapid virological response (RVR), early virological response (EVR) and SVR were achieved in 69.6%, 95.8% and 68.6% of the patients, respectively. Multivariate analysis (odds ratio; 95% confidence interval; P value) indicated age (0.96; 0.93–0.99; 0.012), low baseline viral load (4.65; 2.23–9.66; <0.001) and CC genotype of rs12979860 (7.74; 2.55–23.53; <0.001) but no other SNPs were independent predictors for SVR. In addition, none of the ten SNPs examined were associated with baseline viral load and stages of liver fibrosis. Regarding RVR, low baseline viral load (2.83; 1.40–5.73; 0.004) and CC genotype of rs12979860 (10.52; 3.45–32.04; <0.001) were two critical predictors. As for EVR, only CC genotype of rs12979860 (36.21; 6.68–196.38; <0.001) was the predictor. Similarly, for end of treatment response (ETR), CC genotype of rs12979860 (15.42; 4.62–51.18; <0.001) was the only predictor. For patients with RVR, only low baseline viral load (3.90; 1.57–9.68; 0.003) could predict the SVR. For patients without RVR, only rs12979860 (4.60; 1.13–18.65; 0.033) was the predictor for SVR. Conclusions/Significance rs12979860 is the critical predictor for RVR, EVR, ETR and SVR in treatment-naïve patients of GT1-HCV chronic infection. Furthermore, this SNP is the only predictor for SVR in patients without RVR. These results have provided evidence that rs12979860 is the ideal IL28B SNP for genetic testing in treating patients of GT1-HCV chronic infection.

A delay damage model selection algorithm for NARX neural networks

Recurrent neural networks have become popular models for system identification and time series prediction. Nonlinear autoregressive models with exogenous inputs (NARX) neural network models are a popular subclass of recurrent networks and have been used in many applications. Although embedded memory can be found in all recurrent network models, it is particularly prominent in NARX models. We show that using intelligent memory order selection through pruning and good initial heuristics significantly improves the generalization and predictive performance of these nonlinear systems on problems as diverse as grammatical inference and time series prediction.

A dynamic system approach for radio location fingerprinting in wireless local area networks

This study focuses on the localization using Received Signal Strength (RSS) in dense multipath indoor environments. A dynamic system approach is proposed in the fingerprinting module, where the location is estimated from the state instead from RSS directly. The state is reconstructed from a temporal sequence of RSS samples by incorporating a proper memory structure based on Takens embedded theory. Then, a more accurate state-location correlation is estimated because the impact of the temporal variation due to multipath is considered. An indoor experiment in Wireless Local Area Networks (WLAN) shows the effectiveness of our approach.

Learning a class of large finite state machines with a recurrent neural network

Abstract One of the issues in any learning model is how it scales with problem size. The problem of learning finite state machine (FSMs) from examples with recurrent neural networks has been extensively explored. However, these results are somewhat disappointing in the sense that the machines that can be learned are too small to be competitive with existing grammatical inference algorithms. We show that a type of recurrent neural network (Narendra & Parthasarathy, 1990, IEEE Trans. Neural Networks, 1, 4–27) which has feedback but no hidden state neurons can learn a special type of FSM called a finite memory machine (FMM) under certain constraints. These machines have a large number of states (simulations are for 256 and 512 state FMMs) but have minimal order, relatively small depth and little logic when the FMM is implemented as a sequential machine.

How embedded memory in recurrent neural network architectures helps learning long-term temporal dependencies

Learning long-term temporal dependencies with recurrent neural networks can be a difficult problem. It has recently been shown that a class of recurrent neural networks called NARX networks perform much better than conventional recurrent neural networks for learning certain simple long-term dependency problems. The intuitive explanation for this behavior is that the output memories of a NARX network can be manifested as jump-ahead connections in the time-unfolded network. These jump-ahead connections can propagate gradient information more efficiently, thus reducing the sensitivity of the network to long-term dependencies. This work gives empirical justification to our hypothesis that similar improvements in learning long-term dependencies can be achieved with other classes of recurrent neural network axchitectures simply by increasing the order of the embedded memory. In particular we explore the impact of learning simple long-term dependency problems on three classes of recurrent neural network architectures: globally recurrent networks, locally recurrent networks, and NARX (output feedback) networks.Comparing the performance of these architectures with different orders of embedded memory on two simple long-term dependencies problems shows that all of these classes of network architectures demonstrate significant improvement on learning long-term dependencies when the orders of embedded memory are increased. These results can be important to a user comfortable with a specific recurrent neural network architecture because simply increasing the embedding memory order of that architecture will make it more robust to the problem of long-term dependency learning.