Hyperspectral image (HSI) in Internet of Things (IoT) is a typical small sample data set, which is difficult and costly to label samples manually. In the feature extraction, it is difficult to increase the interclass distance and reduce the intraclass variance according to the limited label information, resulting in easy misclassification of the extracted features. To solve this problem, this article proposes an adaptive pseudo-label feature learning (APFL) model. In the APFL model, a hybrid distance pseudo-label generation (HDPG) method was designed to generate pseudo-labels by iterative multiscale superpixel segmentation using the spectral-spatial mixing distance information, while a pseudo-label feature generation (PFG) method was designed to generate pseudo-label features using pseudo-labels to capture the intraclass average vectors of HSI principal component features. Finally, the extracted pseudo-label features are classified at the pixel level. This APFL model can effectively reduce the intraclass variance and increase the interclass distance of the HSI data, thus improving the interclass separability. We have done comparative verification experiments on five commonly used HSI data sets in IoT. Compared with the current advanced feature extraction methods and classification methods, the proposed APFL model in this article has higher classification accuracy.

Environmental factors that lead to the movement and type of obstacles in and around wireless links are well-known to directly affect channel characteristics. However, while mobile users typically have repeatable daily or weekly patterns with common locations being frequently visited, many protocols along the network stack do not attempt to identify when physical locations are revisited. If wireless channels could be recognized as previously visited, the observance of good and bad decisions in that particular context could dramatically improve some network protocols. In this paper, we present a channel recognition framework which uses the geometrical shape of the link-level performance in a particular context across transmission modes and channel qualities. When attempting to recognize a channel condition, the performance of data transmissions is observed and compared against known channel types to detect similar behavior. The matching channel type can be used as an input to the link adaptation training and resulting decision structure. We perform extensive experimentation on controlled repeatable channels as well as in-field channels to show the validity of the classification algorithm.

Existing rate adaptation protocols have advocated training to establish the relationship between channel conditions and the optimal modulation and coding scheme. However, wireless devices for outdoor and vehicular communications frequently enter environments they have not yet encountered and therefore, have insufficient training for rate adaptation decisions. In addition, protocols are often optimally tuned for indoor environments but, when taken outdoors, perform poorly. In both cases, the decision structure formed offline lacks the ability to acclimate to a new situation on the fly. The diverse and ever-changing environments of increasingly mobile wireless devices call for a rate adaption scheme that can quickly adjust accordingly to form a unique environment set established by the user. In this paper, we propose an on-the-fly, in-situ training (FIT) mechanism which addresses the challenges of making rate decisions with unpredictable fluctuation and lack of repeatability of real wireless channels. We design and conduct extensive experiments on emulated and in-field wireless channels to evaluate the in-situ training process, showing that the rate decision structure can be updated as channel conditions change using existing traffic flows.

In this paper, we develop transmission schemes that maximize the spatial multiplexing rate in multiple-input multiple-output (MIMO) systems with a constraint on the total transmit and processing power. The proposed transmission strategy uses a combination of antenna selection with a bursty transmission strategy to achieve increased rate when complete channel state information is available at the transmitter. We also develop an algorithm, based on dynamic programming, to determine the optimal combinations of antenna configurations and their fractional transmission durations. For a MIMO system with partial channel state information at transmitter (CSIT), we design a feedback scheme, where the information of both the antenna configuration and the precoding matrix is sent to the transmitter. We also provide a heuristic method to design the codebook that sends back antenna configuration information to the transmitter. Numerical results indicate that significant improvements in rate are obtained with the proposed strategy over a traditional precoder feedback strategy.

In this paper, we evaluate the expected rate with quantized feedback over a fading channel with both transmit and computational power constraints. We derive the optimal quantization scheme which can achieve the maximum expected rate in two different scenarios. First, for systems with a short-term power constraint, we derive an algorithm to find the optimum quantization schemes with bursty transmission. Second, for systems with a long-term power constraint, we prove that no explicit bursty transmission is required. We also propose a two-step power allocation method, based on the traditional waterfilling followed by a water adjustment step, which achieves the maximum expected rate. We quantify the gain that the proposed algorithms can achieve for different quantization levels.

An increasing number of adaptive protocols use training data to learn optimal parameter choices for adaptation in wireless communication networks. For instance, several recent papers have studied link adaptation protocols based on context information such as node velocity and SNR. However, a number of embedded sensors providing context information frequently report erroneous values, e.g., GPS errors and accelerometer lag, producing incorrect information about motion. As a result, the relationship between the context information and optimal parameter choices that the adaptive algorithm is attempting to establish is erroneous. In this paper, we propose an outlier detection algorithm, which detects the corrupted information due to system errors. The proposed outlier detection algorithm is based on an alternating minimization approach. To evaluate the performance of the proposed algorithm, we apply it to a link-level context-aware rate adaptation system. Numerical results on emulated channels and in-field testing demonstrate that the proposed algorithm increases the prediction accuracy of the optimal transmission mode by 87% and the throughput by 18%.

Context awareness has received increasing attention with the proliferation of various types of sensors on mobile devices. However, while wireless performance is known to be highly correlated with environmental settings, mobile devices have yet to fully exploit the awareness of context to improve wireless performance. In this paper, we leverage available context information to improve link-level adaptation via decision-tree classifiers and extensively evaluate its performance over emulated channels as well as with in-field trials. We first propose a classification method based on decision trees to select the optimal transmission parameters such as modulation, coding rate and packet size. We then quantify the throughput improvement using the proposed scheme and show that in some scenarios the throughput increases by over 100% compared to traditional SNR-based rate adaptation protocols. Second, we analyze the amount of training to assess the classification scheme. Third, we validate classification-based method by implementation on two different test platforms for extensive experimentation. We reveal the importance of the various contextual attributes used and identify channel type as a key parameter that affects classification performance. Finally, we study and quantify the use of context information across multiple different frequency bands and demonstrate the significant throughput gains that can be obtained.

The application of machine learning algorithms in wireless communications has attracted increasing attention due to the promising performance gains recently achieved. Static classification algorithms have been successfully applied to training protocols that adapt transmission parameters according to context information. However, in reality, there are many time-varying reasons for fading channel quality including mobility of sender, receiver, and/or obstacles within the environment. Moreover, time-varying noise further exacerbates the dynamics of the channel. These problems pose new challenges for the application of static classification algorithms in context-aware algorithms and suggest that sequential classifiers which leverage the temporal dynamics and correlation of context information might be more appropriate. In this paper, we apply sequential training to rate adaptation (ASTRA), leveraging the temporal correlation of context information. In particular, linear and non-linear sequential coding schemes are used in the training process for selecting the modulation/coding rate that achieves the highest throughput for the given context. Experimental results on measurements from emulated and in-field channels demonstrate that ASTRA can significantly increase the accuracy of selecting these target rates by up to 175% and increase the resulting throughput by up to 66% over rate adaptation training which uses static classifier-based methods.