High throughput coherent optical transmitters are critical elements in future optical communication systems, but their performance is constrained by components nonlinearities. We propose a bidirectional recurrent neural network (RNN) based digital pre-distortion (DPD) method, trained using both direct learning (DL) and indirect learning (IL), to effectively mitigate nonlinear distortions. We simulate a 64 Gbaud 64-quadrature amplitude modulation (QAM) coherent optical transmitter under varying levels of nonlinearity and noise. We evaluate the proposed neural network (NN)-DPD against Volterra, look-up table (LUT)-3 and linear DPD solutions. Our results show that the NN-DPD trained via DL consistently outperforms other nonlinear DPD methods, achieving up to a 4.4 dB gain in effective signal-to-noise ratio (SNR) over linear DPD.

This paper presents a novel digital pre-distortion technique leveraging deep reinforcement learning (DRL) to address memoryless nonlinearities caused by nonideal electrical components in communication systems. The proposed DRL based method is shown to outperform traditional approaches—including Volterra series, look-up tables, and machine learning-based indirect learning algorithms—in terms of bit-error rate (BER) and effective signal-to-noise ratio (SNR) improvement. Additionally, the study demonstrates the superior robustness of DRL compared to direct learning architectures (DLA) under abrupt input power variations in power amplifiers. Notably, this work is the first to incorporate transfer learning into the DRL training process, effectively stabilizing performance across varying operating conditions.

This paper investigates an over-the-air federated learning (OTA-FL) system that employs fluid antennas (FAs) at an access point. The system enhances learning performance by leveraging the additional degrees of freedom provided by antenna mobility. We analyze the convergence of the OTA-FL system and derive the optimality gap to illustrate the influence of FAs on learning performance. With these results, we formulate a nonconvex optimization problem to minimize the optimality gap by jointly optimizing the positions of the FAs, the beamforming vector, and the transmit power allocation at each user. To address the dynamic environment, we cast this optimization problem as a Markov decision process and propose the recurrent deterministic policy gradient (RDPG) algorithm. Finally, extensive simulations show that the FA-assisted OTA-FL system outperforms systems with fixed-position antennas and that the RDPG algorithm surpasses the existing methods.

This paper proposes an innovative approach to improve over-the-air federated learning (OTA-FL) systems by integrating fluid antennas (FAs) at the access point. By exploiting the mobility of FAs, we aim to increase the correlation among the users’ channels, thereby improving the learning performance. We analyze the performance of over-the-air computation and the convergence behavior of the OTA-FL system, highlighting the benefits of FAs. Since the learning performance improves as more devices participate in the FL aggregation, we formulate a non-convex optimization problem that maximizes the number of selected users by jointly optimizing FA positions and the beamforming vector, coupled with a user selection policy subject to a mean-squared error constraint. To address environmental dynamics, we describe the problem as a Markov decision process and develop a long short-term memory (LSTM)-based algorithm for efficient decision-making. Simulation results demonstrate that the proposed FA-assisted OTA-FL framework significantly outperforms conventional setups, achieving higher user selection rates and improved learning performance compared to existing benchmarks.

We theoretically compare the performance of optical discrete multi-tone (DMT) and pulse amplitude modulation (PAM) using intensity modulated and direct detection. PAM is a lower cost, lower complexity solution than DMT, however, it is more vulnerable to chromatic dispersion on the C-band. We compare DMT and PAM taking into consideration the interplay of laser linewidth, fiber length, transmission rate, and channel bandwidth. We use a semi-analytical model to examine bit error rates. We study how system parameters shift the performance advantages between DMT and PAM. Our model can also be used to find the best hardware solution and frequency band for a target modulation format and bit error rate.

High speed silicon optical modulators must be compatible with CMOS drivers to reach their full potential for low cost and low power consumption. We designed a CMOS driver for a Mach-Zehnder modulator (MZM) that has been segmented to lower required voltage swings on a per segment basis and to increase the speed. Previous inductor-less drivers produced on the same process, IBM 130 nm CMOS, ran at 10 Gb/s for on-off keying driving a traveling wave (not segmented) MZM. Our chip runs at 20 Gb/s input and performing logic on two data inputs to drive multiple segments to achieve PAM4 modulation, therefore enabling 40 Gb/s operation. The chip provides three drive signals, one for each segment of the MZM. Our measurements verify that at 20 Gb/s the driver provides 2V peak-to-peak voltage swing (4V p-p voltage swing in a push-pull configuration) over a 50Ω load.

In this paper, we analyze the performance of code acquisition system in atmospheric optical code division multiple access (OCDMA) communications using optical orthogonal codes. Memory introduced by temporal correlation of optical fading process precludes us from using the Markov chain model for a code acquisition analysis. By considering this issue, we discuss how to extend the applicability of the Markov chain model to the atmospheric OCDMA communications. We analyze and compare the performance of correlator and chip level detector (CLD) structures in the acquisition system. In our analysis, we consider the effects of free space optical channel impairments, multiple access interference, and receiver thermal noise in the context of semi-classical photon-counting approach. Furthermore, we evaluate the performance of various two stage schemes that utilize different combinations of active correlator, matched filter, and CLD in search and verification stages, and we find the optimum acquisition scheme among them. Numerical results show significant improvement in reducing the acquisition time and required power for synchronization using our optimum scheme in the wireless OCDMA communications.

In this work, we present a clinical prototype with a wearable patient interface for microwave breast cancer detection. The long-term aim of the prototype is a breast health monitoring application. The system operates using multistatic time-domain pulsed radar, with 16 flexible antennas embedded into a bra. Unlike the previously reported, table-based prototype with a rigid cup-like holder, the wearable one requires no immersion medium and enables simple localization of breast surface. In comparison with the table-based prototype, the wearable one is also significantly more cost-effective and has a smaller footprint. To demonstrate the improved functionality of the wearable prototype, we here report the outcome of daily testing of the new, wearable prototype on a healthy volunteer over a 28-day period. The resulting data (both signals and reconstructed images) is compared to that obtained with our table-based prototype. We show that the use of the wearable prototype has improved the quality of collected volunteer data by every investigated measure. This work demonstrates the proof-of-concept for a wearable breast health monitoring array, which can be further optimized in the future for use with patients with various breast sizes and tissue densities.

We experimentally demonstrated the transparent convergence of distributed antenna system in a digital wavelength division multiplexed passive optical network system employing reflective semiconductor optical amplifier (RSOA) at the remote antenna unit (RAU). The spurious free dynamic range and frequency response of the RSOA is measured to meet WiFi's requirements. RSOA's electrical-to-optical conversion efficiency improves significantly by down converting the WiFi signal to intermediate frequency band. By incorporating the up/down mixer in Central Office and RAU, we experimentally achieve full-duplex WiFi transmission along with 1 Gb/s downlink digital on-off-keying signal in single mode fiber up to 15 km (64QAM), 40 km (16QAM), and 60 km (QPSK) with 1 meter wireless path working in half-duplex mode. The bit-error-rate below the forward error correction threshold (2 × 10-3) is observed in all cases whereas the 1 Gb/s digital downlink signal is error-free up to 20 km of fiber.

This paper presents a 2.4 GHz low-power CMOS On-Off Keying receiver front-end which contains a low noise amplifier with a novel down conversion mixer that is designed for wireless forward telemetry link in neural stimulation applications. The transceiver operates between 2.4 and 2.5 GHz to support the full industrial, scientific and medical band. The post-layout simulation results show that the fully integrated low-noise amplifier exhibits a gain of 15 dB, a noise figure of 1 dB at 2.4 GHz, and the input matching (S11) is -16 dB. The proposed mixer is resistor-less and designed based on current starved delay elements in a Gilbert topology. The transceiver implemented in a TSMC 0.18 μm CMOS technology uses a supply voltage of 1.2 V, supports a data rate of up to 100 Mbps, and consumes 7 mW. The total size of the proposed receiver front-end equals 0.38 mm2.

We examine the packet loss performance of scalable all-optical routers exploiting fiber delay lines, under non-uniform and bursty data center traffic conditions. Our analysis reveals that a load balancing stage, capable of evenly distributing the incoming traffic in both space and time, is indispensable in wavelength-routing data centers with a distributed-buffer architecture.

UWB impulse radio transmissions are power limited as they must respect a frequency mask with low total permitted power and severe isolation requirements for the GPS band. Shaping of the short UWB impulse maximizes transmit power subject to these constraints. We have previously published results for intricate pulse shaping in the optical domain that achieves record levels of power transmission. The most successful pulse shaping techniques in the electrical domain have been achieved with derivatives of the Gaussian pulse shape. The Gaussian monocycle is a low performance choice in terms of transmit power, but nonetheless the subject of intense research activity. In this paper we compare and contrast these UWB pulses, and quantitatively compare their performance in realistic systems (specific RF and antenna hardware). There is the perception that failings in the pulse shape (especially for the monocycle) can be compensated by simple highpass filter and the bandpass nature of the UWB antenna. We fabricate and characterize two UWB antenna designs available in the literature, as well as one commercial antenna. We use our optical pulse shaping device to experimentally investigate the three different pulse shapes and measure their transmitted power spectral density with each of the UWB antennas. We find that the monocycle is significantly less powerful than the others. The fine pulse shaping available in optical processing provides 1.7 to 2.9 dB gain over the electrical processing methods, depending on the antenna used.

All-optical switching has been proposed to overcome the limitations of electronic switches in terms of scalability, speed, footprint, and power consumption. A key passive optical component to bypass electronic processing limitations is the arrayed waveguide grating (AWG). Switch architectures combining wavelength converters and fiber delay lines provide tunable routing and contention resolution when used with AWGs. An AWG passively routes either single or multiple input port wavelengths to its output ports. A single wavelength per port strategy reduces crosstalk within the AWG, but drastically increases the dimensionality of the device. Physical constraints on AWG design limit the port number for the foreseeable future to under 100. To scale optical switches to emerging network requirements, we can use multiple wavelengths per port. In this paper we examine one multiple wavelength per port architecture and quantify the physical layer impairments due not only to the AWG crosstalk, but also Q-factor degradation due to multiple wavelength conversions, and as a function of the number of recirculations in the contention resolution delay lines. While previous work has addressed this issue in terms of accumulated loss, we focus on accumulated relative intensity noise and amplified spontaneous emission.

We propose and experimentally demonstrate generation and distribution of optimally power efficient impulse-radio ultra-wideband (IR-UWB) waveforms over passive optical networks (PONs) for high bit rate wireless data distribution in the home. We use a reconfigurable UWB pulse generator in the context of radio over fiber delivery over a PON for fiber to the home/premises services. The single mode fiber PON feeder is used to create the required dispersion for the frequency-to-time conversion for optical pulse shaping. As the feeder length varies for different PONs, a tunable UWB pulse generator compensates for variable dispersion. We can compensate the effect of feeder length over a high dynamic range. We experimentally demonstrate the generation of the same UWB pulse for feeder lengths of 17, 27.5 and 35.5 km by tuning the optical pulse generator.

Traditionally the Gaussian monocycle pulse and its higher derivatives have been proposed and deployed as impulse radio ultra-wideband (IR-UWB) pulses. Although relatively easy to generate in electronics, these pulses are ill-adapted to the sharp cutoffs in the US Federal Communications Commission (FCC) mask. The combination of these pulses with passband UWB antennas with steep roll-off improves power efficiency vis-a-vis the FCC mask. This approach is still quite sub-optimal and pulse shaping can provide marked improvement (to 3 dB) over the best traditional combinations. We show that optimal design of UWB waveforms, when taking into account antenna gain profiles, improves the power efficiency of the pulses. Three typical antennas are considered. A nonlinear optimization process is used to design an efficient pulse for each antenna. The proposed optimization is based on the hybrid genetic algorithm and a sequential quadratic program. We demonstrate that this method finds efficient pulses under severe antenna distortion. Simulation results confirm that the optimally designed pulses have superior performance compared to the more common Gaussian monocycle and the Gaussian fifth-derivative pulse.

We analyze the performance of fiber fault monitoring of a PON using a centralized, passive optical coding (OC) system. We develop an expression for the detected monitoring signal, study its statistics, and contrast our OC monitoring system with standard optical code division multiplexing (OCDM) data communication. We derive a new closed form lower bound expression for the interference probability, and use this to find the signal-to-interference ratio (SIR). A service provider cannot control the physical layout of homes in a coverage area; measuring performance for one network layout would be insufficient to test our proposed monitoring system. Client geographic distribution has a significant impact on the SIR. We consider five different PON geographical distributions and study their effect on OC monitoring performance. Our results show that SIR is sufficient to successfully monitor the network. In addition, we find that the uniform radial (UR) distribution, an analytically tractable distribution, gives good performance estimation and can therefore be a useful tool in characterizing performance in terms of both SIR and signal-to-noise ratio (SNR).

We evaluate the theoretical performance of recently proposed optical coding (OC) technology for fiber fault monitoring of a PON through the signal-to-noise ratio (SNR), the signal-to-noise-plus-interference ratio (SNIR), and the falsealarm probability. First, we develop a mathematical model and expressions for the detected monitoring signals considering a square law detector and using realistic parameters. Second, we address the effect of the transmitted pulse power, network size and light source coherence on the performance of both one-dimensional (1D) and two-dimensional (2D) OC monitoring systems. We show that the transmitted pulse width can be optimized to trade-off the interference and the detection noises. We give simple analytic equations for this optimal pulse width as a function of network parameters. Both 1D and 2D coding schemes are considered. We find that, under perfect dispersion compensation, an incoherent source performs better than lasers for 1D coding. In addition, 2D coding using lasers offer very good performance and supports networks up to 128 customers with SNIR¿10dB; a promising candidate for future high capacity PON. Finally, we apply Neyman-Pearson testing to the receiver of our monitoring system and investigate how coding and network size affect the operational expenses (OPEX) of our monitoring system.

The Multicanonical Monte Carlo (MMC) technique is a new form of adaptive importance sampling (IS). Thanks to its blind adaptation algorithm, it does not require an in-depth system knowledge for exploitation as does traditional IS. Hence MMC is a practical, handy tool to estimate via simulation the probability of rare events in complex telecom systems, such as the symbol error rate or the outage probability. In this paper, we present the analytical connections between MMC and IS, and describe the recursive algorithm via which MMC seeks an optimal "flat-histogram" warping. We also provide practical guidelines on how MMC can be successfully applied in telecom to achieve accelerations of simulation time by many orders of magnitude with respect to standard Monte Carlo.

We study the impact of the shape of optical filters on the bit error rate (BER) of a multi-channel spectrum-sliced wavelength division multiplexed (SS-WDM) system, incorporating nonlinear semiconductor optical amplifiers (SOA) to suppress the excess intensity noise, using a parallel implementation of Multicanonical Monte Carlo (MMC) algorithm. Both the slicing filter (SF) at the transmitter, and the channel selecting filter (CSF) at the receiver are allowed to be independently chosen from a reduced set of filter shapes. BER curves corresponding to a fixed SF bandwidth and bit-rate are traced for various bandwidths of the CSF. Both the single channel case, where the only impairment is the "filtering effect", and the multi-channel case, with crosstalk, are considered. We estimate the achievable spectral efficiency of SOA-assisted SS-WDM passive optical networks (PON) when both SOA-based intensity noise suppression, and forward error correction (FEC) are employed.

In this paper we investigate both experimentally and via simulation the monitoring of fiber link quality in a PON using optical coding technology. We use a new, simple and cost-effective coding scheme well adapted to the monitoring application, namely periodic coding. We discuss design issues for periodic coding and the optimal detection criteria. We develop a reduced complexity maximum-likelihood sequence estimation (RC-MLSE) algorithm for monitoring. We conduct experiments to validate our detection algorithm using four periodic encoders that we designed and fabricated. These encoders were placed at roughly equal distances (within a meter) to represent a partial return from a very high density (geographically) PON. The measured data were fed into our detection algorithm and the exact location of each subscriber was correctly identified. Using the experimental data for the encoder's impulse responses, we completed Monte-Carlo simulations for more realistic PON geographical distributions with randomly located customers. Error-free detection is achieved. We also highlight the importance of averaging to remedy the power/loss budget limitations in our monitoring system to support higher network sizes.

In this paper we study the performance of source matching technique for an optical code division multiple access (OCDMA) system in the presence of source coherence effects and square law detection process. We use a binary-asymmetric channel (BAC) model for an OCDMA system employing an all-optical passive correlator receiver. Source coherence effects lead to relative-intensity-noise (RIN) and phase-induced-intensity-noise (PIIN), which are included in our analysis. Previous studies only considered multiple access interference (MAI) noise, resulting in a Z-channel (where errors only occur for the transmission of data bit zero) model, and neglected RIN and PIIN. The presence of RIN and PIIN leads to errors occurring for both transmitted data bits one and zero, thus a BAC model. We show that source matching gain depends on the normalized source coherence time, defined as the ratio of the optical source coherence time to the bit duration. Our analysis shows that, while MAI limited analysis predicts that increasing the number of users increases the source matching gain, when taking into account RIN and PIIN, source matching gain is both bit rate and source type dependent, and tends to zero for very high numbers of users.

We propose a new semi-analytic method for the study of the performance of a RAKE receiver in a direct-sequence code-division multiple-access (DS-CDMA) based ultra wideband (UWB) system. This method is intended as a tool to facilitate further performance studies of this type of system and, ultimately, to lead to a fully analytical model. We also analyze system performance, in the presence of intersymbol interference (ISI) and narrowband interference (NBI), using two RAKE-based receivers. The first receiver is the classical RAKE, which employs maximal ratio combining (MRC). In the second receiver, MRC combining is replaced by linear minimum mean-squared error (MMSE) combining.

We present an experimental demonstration and simulation results of frequency encoded code division multiple access (FE-CDMA) using a broadband source, fiber Bragg gratings for encoding/decoding of unipolar m-sequence codes and balanced detection. The bit error rate is measured for up to four simultaneous users, at 155 Mb/s and 622 Mb/s. Exploiting the excellent match between simulation and experiment, we conclude with a prediction of potential system capacity of an optimised FE-CDMA system.

This paper reports for the first time an all-optical wavelength conversion of an incoherent optical signal to a coherent optical signal using the cross-gain modulation in semiconductor optical amplifiers. Wavelength division multiplexed networks using broadband sources have performance limited by beat noise; conversion from a noncoherent to a coherent signal eliminates beat noise and enables the elimination of the bit error rate floor.

We demonstrate the effectiveness of multiuser detection for an ultra-wideband (UWB) pulse based direct sequence spread spectrum system using code division multiple access. Extensive simulations were run using channel soundings of the 2-8 GHz band collected in a residential setting and characterized by a high level of multipath fragmentation. We show that the adaptive minimum mean square error (MMSE) multiuser detection (MUD) receivers are able to gather multipath energy and reject intersymbol and interchip interference for these channels to a much greater extent than RAKE receivers with 4 and 8 arms. We also demonstrate the adaptive MMSE is able to reject a narrowband IEEE 802.11a OFDM interferer, even for signal-to-interference ratio as severe as -30 dB. We show the adaptive MMSE exhibits only a 6 dB penalty relative to the single user case for the heavy multi-access interference (number of asynchronous users equal to spreading code length). The practical RAKE receivers were incapable of effectively rejecting either the strong narrowband interference or the heavily loaded wideband interference. Even more moderate levels of interference caused significant degradation in the performance of the practical RAKE receivers.

We propose a new multirate optical communication system using optical fast frequency hopping CDMA (OFFH-CDMA) for multimedia applications in which different quality of services (QoS) are required. In this system, each user needs only to transmit the minimum required power to achieve a desired signal to interference ratio (SIR). We assign different power levels to each rate through an average interference-based power control algorithm using variable optical attenuators. Such an approach minimizes interference and at the same time provides variable QoS constraints for different traffic types. The simulation shows a great improvement in the system capacity.

Sun et al. (see Electron. Lett., vol.32, p.1490-91, Aug. 1996) reduced the set of coupled first order nonlinear partial differential equations determining the wavelength-dependent, time-varying doped-fiber amplifier gain into a single ordinary differential equation (ODE). In this paper we further simplify the ODE, greatly enhancing its utility as an analysis and design tool. We find that the gain dynamics are completely specified by the total number of excited ions. We demonstrate that channel addition causes much faster transients than channel dropping in wavelength division multiplexing networks. We approximate the solution of the ODE by an exponential with the time constant given as a function of the amplifier parameters.

We propose fast optical frequency-hop code division multiple access (FH-CDMA) for high bandwidth local area networks. Encoding and decoding are achieved by tunable fiber Bragg gratings. Frequency hopping offers many advantages compared to the previously proposed optical CDMA multiple access techniques. The simultaneous utilization of the time and frequency domains offers notable flexibility in the selection of codes, however code families previously developed for radio frequency (RF) communications are not directly applicable to an optical FH-CDMA system. We propose codes to meet the special constraints imposed by our encoding device and present theoretical and simulation results for their performance.

Support of multiple data services can be accomplished via code division multiple access (CDMA) with a constant chip rate, but variable data rates. We present exact equations for the cross-correlation of codes in a multiple-data rate system, and thus equations for the multiple access interference (MAI) and bit error rate (BER). Previous analysis has found the EER using the mean of the MAI assuming codes are truly random sequences. We show that Gold and Kasami sequences have a cross-correlation with a distribution closely approximated by a Gaussian one, and present confidence intervals to quantify the performance under the random code assumption.

We address the problem of suppression of a digital narrow-band interferer in direct-sequence spread spectrum (DSSS) communications. We focus on the adaptive suppression method proposed by Honig, Madhow and Verdti (see IEEE Trans. Inform. Theory, vol.41, p.944-960, 1995) for wide-band interference, applying it to a narrow-band interferer. We identify the eigenspaces of the system dynamics to analyze the convergence of the adaptive version of the minimum mean square error (MMSE) algorithm for this application. Using this subspace approach we are able to: (1) significantly decrease the convergence times via a new constraint on the step size in adaptation; (2) introduce a simple parameterization of the mean output energy (MOE) and signal-to-interference ratio (SIR) to compare performance of various receivers; and (3) identify modes of operation where the algorithm will cease to effectively cancel interference. We propose a new adaptive receiver that avoids the convergence anomalies identified, while capitalizing on the new step size for faster convergence. Simulation results to support theoretical results are presented.

Code division multiple access (CDMA) has been proposed for use in fiber high speed point-to-point systems. Previous research into CDMA has centered on completely coherent detection, that is, when the phases of all users in the system are known or tracked. While coherent detection is a reasonable assumption in many radio frequency systems, semiconductor lasers suffer from serious phase variation that makes tracking of all phases difficult to accomplish. This paper examines the effects of phase drift on several multiuser detectors for binary phase shift keying (BPSK), as well as two detectors for binary frequency shift keying (BFSK). It is shown that when the desired user's phase can be tracked, BPSK match filtering offers better bit error rates than BFSK, and closed form solutions for the asymptotic error probabilities are derived. When there is no phase tracking, it is shown that the probability of error of the multiuser detectors approaches one half. The analysis is based on a proof of the asymptotic normality of the interference caused by even a single undesired user. All asymptotic results are for infinitely long spreading codes that can be modeled as a random sequence of plus and minus ones. Whether or not the desired user's phase is tracked, all receivers examined suffer from the near-far problem.>

Due to the scarcity of frequency allocations for emerging personal communications services, it has been proposed that spread spectrum (SS) signals be used in such services so that they can be overlaid on existing frequency band occupants. Such SS signals would be constrained in power so as not to interfere with pre-existing, narrowband users. In view of this application, suppression of strong narrowband signals that interfere with commercial SS communications systems is being investigated with renewed interest. Previous research has modeled narrowband users as sinusoidal or autoregressive signals, and consequently has advocated use of predictive altering to suppress such signals. In the present paper it is shown that when the narrowband interferer is in fact a digital communications signal these methods are less effective than techniques derived from multiuser communications. In particular, by modeling a narrowband user as a digital signal, optimal and asymptotically (low background noise) optimal linear algorithms for the recovery of the SS bit stream are developed. This technique represents not only a promising new approach to narrowband interference suppression, but also a new application for multiuser detection theory.>

Spread spectrum (SS) communications offers a promising solution to an overcrowded frequency spectrum amid growing demand for mobile and personal communications services. The proposed overlay of spread spectrum signals on existing narrowband users implies strong interference for the SS system. The paper discusses how system performance can be improved by preprocessing to suppress narrowband interference. Linear prediction filters have been proposed since the 1980s for suppression of narrowband interference. Vijayan and Poor (1991) proposed nonlinear methods of suppressing the narrowband signal with significant increase in the SNR improvement. The present authors derive an enhancement to this nonlinear prediction and achieve further improvement by applying the technique to interpolating filter structures. Finally, they extend results to the case of multiple spread spectrum users and demonstrate how nonlinear filtering can dramatically outperform linear filtering.>

In 1990 Vijayan and Poor proposed nonlinear predictive methods for suppressing narrowband interference signals in spread spectrum (SS) with significant increase in the SNR improvement. This paper describes a further enhancement of their nonlinear adaptive prediction algorithm, and introduces a nonlinear version of the interpolating filters proposed in previous studies.>