One-class-classification remains an important problem in machine learning, which is related to data representation and outlier detection, but different from them in several aspects. In the present contribution we propose an one-class-classifier based on a prototype vector quantization model. We modeled a corresponding cost function to account for aspects of representation learning and to appropriately evaluate the one-class classifier. The prototype-based model ensures a local representation of the target class. After this introduction, we obtain an interpretable one-class classifier model. We demonstrate the capabilities of the approach by applying the classifier to illustrative toy data examples as well as on real data in a medical context.

Abstract In this contribution, we consider the classification of time series and similar functional data which can be represented in complex Fourier and wavelet coefficient space. We apply versions of learning vector quantization (LVQ) which are suitable for complex-valued data, based on the so-called Wirtinger calculus. It allows for the formulation of gradient-based update rules in the framework of cost-function-based generalized matrix relevance LVQ (GMLVQ). Alternatively, we consider the concatenation of real and imaginary parts of Fourier coefficients in a real-valued feature vector and the classification of time-domain representations by means of conventional GMLVQ. In addition, we consider the application of the method in combination with wavelet-space features to heartbeat classification.

In this work, we consider a joint grouping and scheduling algorithm for fair user scheduling in massive multiple-input multiple-output (MIMO) systems that support Internet of Things (IoT) application scenarios. Our proposed method serves the users consecutively in groups, where the group sizes are derived from the optimal number of active users in the zero-forcing downlink channel. To select the best group members we adopt a version of the popular semi-orthogonal user selection (SUS) algorithm. Specifically, we identify the drawbacks of the original SUS algorithm when operating in a massive MIMO system and present a modified SUS strategy (SUS-M) for user selection. Simulation results demonstrate that our proposed method outperforms the conventional SUS strategy and provides improved fairness in terms of the individual user rates.

In this paper, we develop receiver concepts for transmission with space-time block codes (STBCs) over frequency-selective fading channels. The focus lies on Alamouti's space-time block-coding scheme, but the results may be generalized to other STBCs as well. We show that a straightforward combination of conventional equalizers and a space-time block decoder is only possible if at least as many receive antennas as transmit antennas are employed, but not for the practically interesting case of pure transmit diversity, for which space-time coding had been originally developed. This restriction is circumvented by our approach. Here, the structural properties of the transmit signal of space-time block coding, which is shown to be improper (rotationally variant), are fully used. For this, equalizers with widely linear (WL) processing are designed, such as a WL equalizer, a decision-feedback equalizer with WL feedforward and feedback filtering, and a delayed decision-feedback sequence estimator with WL prefiltering. Simulation results demonstrate that the proposed concepts may be successfully employed in an enhanced data rates for GSM evolution (EDGE) receiver, especially for pure transmit diversity. Here, significant gains can be observed, compared with a conventional single-input single-output transmission.

We propose several equalization schemes based on widely linear processing (WLP). The received signal and its complex conjugate are separately filtered and the results are linearly combined. It is shown that WLP yields a gain in performance if the (noiseless) received signal can be interpreted as the convolution of a real-valued data sequence and an equivalent complex-valued intersymbol interference channel impulse response. Such a model applies to, e.g., amplitude-shift keying, offset quadrature amplitude modulation, and binary minimum-shift keying-type modulation. We consider receivers without and with decision feedback. Finite impulse response filters are derived for these structures, which are optimum with respect to the zero-forcing and minimum mean-squared error (MMSE) criteria, respectively. In the MMSE case, adaptive algorithms for filter adjustment are given. Infinite filter orders are investigated in order to obtain analytical performance results. Furthermore, suboptimum trellis-based detection with widely linear preprocessing is briefly discussed. It is demonstrated analytically and by numerical examples that widely linear schemes may outperform conventional schemes significantly, depending on the considered application.

In this paper, we develop receiver concepts for transmission with space-time block codes (STBC) over frequency-selective fading channels. The focus lies on Alamouti's STBC, but the results may be generalized to related STBC. We show that a straightforward combination of conventional equalizers and a space-time block decoder is only possible if at least as many receive antennas as transmit antennas are employed, but not for the practically interesting case of pure transmit diversity. This restriction is circumvented by our approach. Equalizers with widely linear (WL) processing are designed, utilizing the structural properties of the transmit signal of space-time block coding, which is shown to be improper (rotationally variant). These schemes are especially suited for equalization of high-level modulated signals, which are used in third-generation time-division multiple access mobile communications standards such as EDGE (Enhanced Data Rates for GSM Evolution).

A novel multiuser detector for direct sequence code division multiple access is proposed. The receiver performs iterated soft decision interference cancellation (ISDIC) based on multiuser interference suppression filters designed for minimization of the mean-square error. Assuming a complex modulation format, we show that the multiuser interference becomes rotationally variant in the course of the iterations. Regarding this rotational variance in the design of the multiuser interference suppression filter, the presented iterative multiuser detector achieves significant performance gains compared with conventional ISDIC employing a standard minimum mean-squared error filter which is optimum only for rotationally invariant multiuser interference.

A practically interesting approach for iterative channel estimation, multiuser detection, and single-user decoding based on maximum a posteriori symbol-by-symbol estimation for direct sequence/code-division multiple-access (DS-CDMA) is proposed. The receiver relies on the output of a bank of matched filters for each user and each path, and combines interference cancellation with iterated soft-decision feedback to improve channel estimation accuracy and data symbol reliability in course of a few iterations. We show that in this way, near single-user channel phase and amplitude estimation accuracy is achieved for frequency-selective fading channels, even in highly loaded systems, and illustrate that reliable data symbol estimation can be performed.

A novel robust noncoherent receiver for minimum mean-squared error (MMSE) interference suppression for direct-sequence code-division multiple access (DS-CDMA) is proposed. The receiver consists of a linear MMSE filter and a decision-feedback differential detector (DF-DD). The performance of the proposed scheme is investigated analytically and by computer simulations. It is shown that the loss compared to coherent MMSE interference suppression is limited and can be made arbitrarily small by increasing the observation window used for calculation of the reference symbol of the DF-DD. Hence, the regarded noncoherent receiver is near-far resistant. For adjustment of the MMSE filter coefficients three noncoherent adaptive algorithms are proposed. In contrast to coherent adaptive algorithms, these noncoherent algorithms have the important advantage that they also converge if the channel phase is time-variant.

In this paper, a novel robust noncoherent receiver for MMSE interference suppression for DS-CDMA is proposed. The receiver consists of an MMSE filter and a decision-feedback differential detector (DF-DD). The performance of the proposed scheme is investigated analytically and by computer simulations. It is shown that the loss compared to coherent MMSE interference suppression is limited and can be made arbitrarily small by increasing the observation window used for calculation of the reference symbol of the DF-DD. For adjustment of the MMSE filter coefficients a noncoherent normalized least-mean-square (NC-NLMS) algorithm is proposed which is robust against channel phase variations.

We give the asymptotic signal-to-interference ratio for the so-called widely linear minimum mean-squared error multiuser receiver for transmission of multiple users to a common receiver with DS-CDMA. Observing that widely linear multiuser detection virtually doubles the spreading factor, we show that transmission with real-valued channel symbols can lead to a higher spectral efficiency than with complex-valued channel symbols when each user employs its own complex-valued random spreading sequence.