Abstract Pooling operations, such as average pooling, strided convolution, and max pooling, have become fundamental components of convolutional neural networks (CNNs) due to their ability to capture local features, expand receptive fields, and reduce computational costs. However, in the context of semantic segmentation, these pooling techniques can lead to the loss of crucial spatial details that are necessary for accurate pixel-level predictions. To tackle this issue, extensive research has focused on refining deep CNN models through architectural adaptations and novel training methods. Recent studies have demonstrated the importance of pooling layers, exemplified by innovations like the introduction of wavelet pooling. In our study, we highlight the value of incorporating our previously proposed matched wavelet pooling (MWP) into CNNs to enhance semantic segmentation pipelines. The core concept of MWP challenges the notion that including all sub-bands generated from wavelet decomposition consistently improves accuracy. Instead, we advocate for selecting specific sub-bands for the pooling process in each image during both training and testing. This approach introduces sub-band selection protocols customized for image-specific pooling, designed specifically for semantic segmentation CNN architectures, with a particular focus on the UNet and SegNet models. Across three widely used datasets, our proposed MWP- based pipeline, featuring the MWP-UNet architecture, consistently outperforms conventional pooling methods. It achieves a significant average improvement in intersection over union (IoU) of over 25% compared to recent literature. Additionally, our MWP-SegNet model outperformed the standard SegNet by 12.5% mIoU, further demonstrating the effectiveness of our matched wavelet pooling approach across different network architectures.

Abstract In this paper, the generation of new dragon fractal shapes with chaotic iteration parameters is introduced as the main component of a new efficient approach for different cryptographic applications. This process involves applying a chaotic map, which is considered the initiator pattern, to generate different chaotic dragon fractal (ChDrFr) shapes in lieu of lines (which are classically used to generate dragon fractals). This is the new concept of this paper. The used chaotic maps are sensitive to their initial conditions and are characterized by randomness; hence, the resulting scheme is highly secure. As the resulting ChDrFr shapes have sparse structures, the spaces are packed with random values generated from another 5D hyper chaotic map. For encryption applications based on the substitution approach, one of the five generated ChFrDr shapes can be used to construct a chaotic fractal (ChFr) S-Box, while the other four ChDrFr shapes can be used for diffusion purposes. As an application to these new ChDrFr shapes and the ChFr S-Box, we introduce in this paper a new highly secure image encryption algorithm. A Henon chaotic map is used as the initiator of the ChDrFr shapes. The integer wavelet transform (IWT) is used to generate an approximation and three detail sub-bands for the original image. As the approximation sub-band contains a considerable amount of information about the original image, the above-described ChFr S-Box is used as a replacement for each pixel’s value in this sub-band. Then, the resultant substituted image is diffused with one of the generated ChFrDr shapes. The other three ChDrFr shapes are XORed with the details sub-images. Numerical simulation is applied to ensure the efficacy of encrypted images against different attacks. In particular, the correlation coefficient between the initial and the generated images is shown to be nearly zero. Moreover, tests reveal that the information entropy of the encrypted images and UACI were close to their optimum values. The properties of the newly proposed ChDrFr-based encryption algorithm are compared to the ones obtained by other encryption algorithms, and the results prove the superiority of this newly proposed algorithm to other types of encryption methods.

In this paper, we propose two novel methods for downlink preamble detection and initial coarse time synchronization for mobile WiMAX systems and other OFDM Systems that adopt similar preamble structures. The two methods are based on frequency domain processing. The two methods are compared to the conventional time domain method for frame detection in WiMAX and to another frequency domain algorithm in the literature. The proposed methods are less complex while providing similar or superior performance in the SNR range of interest. The performance measures are the detection and false alarm probabilities, in addition to the Mean Acquisition Time which is evaluated for all the considered methods. Since in the initial state the mobile station is in the state of searching for the downlink preamble, we take into account the interfering effect from other nearby Mobile Stations transmission in the uplink subframe with power larger than the received power from the Base Station in the downlink subframe. Simulation results show the robustness of the new proposed methods in different channel conditions and interference environments.

The selective mapping (SLM) scheme is one of the most popular peak-to-average power ratio (PAPR) reduction techniques proposed for multi-input multi-output orthogonal frequency division multiplexing (MIMO-OFDM) systems. One of the major disadvantages of this scheme is the need for the transmission of side information (SI) bits to enable the receiver to recover the transmitted data. The authors present a small overhead SLM (s-SLM) scheme for space–time block coded (STBC) MIMO-OFDM systems. This proposed scheme improves the system bandwidth efficiency and achieves a significantly lower bit error rate (BER) than the individual SLM (i-SLM) and direct SLM (d-SLM) schemes. In addition, approximate expressions for the complementary cumulative distribution function (CCDF) of the PAPR and the average BER of the proposed s-SLM scheme are derived. The simulation results show that the proposed s-SLM scheme improves the detection probability of the SI bits and hence gives a better performance than the i-SLM and the d-SLM schemes.

Presents a simple continuous analog hardware realization of the random neural network (RNN) model. The proposed circuit uses the general principles resulting from the understanding of the basic properties of the firing neuron. The circuit for the neuron model consists only of operational amplifiers, transistors, and resistors, which makes it candidate for VLSI implementation of random neural networks with feedforward or recurrent structures. Although the literature is rich with various methods for implementing the different neural networks structures, the proposed implementation is very simple and can be built using discrete integrated circuits for problems that need a small number of neurons. A software package, RNNSIM, has been developed to train the RNN model and supply the network parameters which can be mapped to the hardware structure. As an assessment on the proposed circuit, a simple neural network mapping function has been designed and simulated using PSpice.

This paper presents a novel random neural network (RNN) based soft decision decoder for block codes. One advantage of the proposed decoder over conventional serial algebraic decoders is that noisy codewords arriving in non-binary form can be corrected without first rounding them to binary form. Another advantage is that the RNN, after being trained, has a simple hardware realization that is ideal for implementation as a VLSI chip. The proposed decoder is tested on Hamming linear codes and the results are compared with that of the optimum soft decision decoder and the conventional hard decision decoder. Extensive simulations show that the RNN based decoder reduces the error probability to zero in the range of the error correcting capacity of the used code. On the other hand, it is much better than the hard decision decoder for codewords corrupted with more errors.

The performance of GMSK/DS spread-spectrum signaling with binary phase-shift-keying and two-bit differential detection operating in a fast Rayleigh frequency dispersive mobile channel is investigated. A generalized channel model, that takes into consideration the presence of faded cochannel interference signals, is used for analysis. Closed form expressions are derived for the average error probability and are numerically evaluated for many channel conditions. The results clearly show that the incorporation of DS signaling, with relatively small processing gain, highly improves the performance of GMSK in severe channel degradation environments.

A new technique of interference cancellation for synchronous optical CDMA communication systems is proposed We utilized the properties of modified prime sequences under the assumption of a Poisson shot noise model for the PIN photodiode and synchronization between users' symbols. The interference cancellation technique is based on estimating the multiuser interference (MUI) through the special correlation properties of the used codes. The estimated MUI is then used to adapt a threshold level to optimally detect the concerned user's information and to suppress the MUI. Detailed derivations of the closed forms for the bit error rate are presented for both cases of with and without the suggested MUI suppression. Numerical results clearly show the advantages of the proposed technique and proves that it outperforms a system without interference cancellation.

The use of pre-RAKE combing technique decreases the complexity of the portable unit in time division duplex (TDD) applications while achieving the same multipath diversity effect of the RAKE receiver. Our analysis is focused on three main issues. First a comparison of the various diversity techniques and studying the effect of the diversity order of three different diversity combining techniques. Second, studying the effect of imperfect channel estimation on the system performance and comparing it with the ideal pre-RAKE system. The obtained conditional probability of error is evaluated through Monte-Carlo integration. Third, a comparison between pre-RAKE and RAKE systems under the same conditions. Both random codes and orthogonal codes are considered in investigating the performance of the two systems.

This paper studies the use of pre-RAKE combining technique in code-division multiple-access (CDMA) systems with QPSK carrier modulation and time division duplex (TDD) in wide-band wireless portable communication systems. The system performance is analyzed in a generalized Rayleigh fading multipath channel. Our analysis is focused on two main issues. The first issue is comparing the various diversity techniques and studying the effect of the diversity order. Two diversity combining techniques, namely, maximal ratio combing (MRC) and equal gain combing (EGC) supplemented by a selection diversity for the strongest paths in gain are examined. The second considered issue is studying the effect of imperfect channel estimation on the system performance and comparing it with the ideal QPSK pre-RAKE system.

The blind equalization of data distorted in dispersive channels with considerable ISI is investigated. The blind channel estimation is realized without using training sequences. Two methods based on the higher order spectra and polycepstra are considered. The estimated channel response is used to adaptively equalize the distorted data. The speed of convergence of the adaptive equalizer is highly increased by suitable filtration of the estimated channel impulse response before adaptive equalization. This is done by providing smart initial weights for the adaptive equalizer through inverse filtering or matched filtering of the estimated channel response. The addition of a deconvolving filter to the matched filter is suggested to cancel the effects of the high offset correlation values of the channel impulse response that is characteristic of channels with severe ISI.

A soft decision decoder of linear block codes based on artificial neural networks (ANN) is presented. Various structures of the ANN decoder (ANND) are introduced and compared with the optimum soft-decision decoding (SDD) and hard-decision decoding (HDD) methods. Although the performance of the SDD is still optimum, the massive parallel nature of the neural network is shown to be more suitable to high data rate communication systems for which optimum decoding is impractical because of complexity and time consumption. The proposed ANND bridges the gap in performance and complexity between HDD and SDD. It is shown that the proposed receiver has superior performance in noisy channels with low signal to noise ratios.

Matched chirp modulation (MCM) is an improved version of chirp modulation (CM) in which the frequency sweep rate is matched to the dispersion parameters of the communication channel. It is an attractive constant-envelope digital modulation technique for adaptive transmission over dispersive communication channels. The performance of the optimum noncoherent (NC) and partially coherent (PC) receivers for MCM is analyzed and evaluated. In particular, the optimum value of the time-bandwidth product is evaluated, and the probability of error is calculated for the optimum NC and PC receivers.>

The feeding of loudspeaker arrays according to special codes of the Huffman type in order to obtain almost omnidirectional intensity patterns is investigated in this paper. These codes are nonbinary and only six discrete levels are allowed to simplify the array design. The corresponding intensity patterns are shown to result in much better omnidirectional patterns than those resulting by feeding with Barker codes or other trial and error binary codes that were introduced recently. The results have direct applications in underwater arrays, public address systems and accoustical measuring systems.

The feeding of loudspeaker arrays by special sequences, or codes, with the purpose of obtaining isotropic radiation intensity patterns is considered in this paper. In particular, new codes of the Huffman codes type, which are generated by combination of Barker codes, are considered. This type of feeding is shown to result in almost isotropic patterns which are superior to those obtained by feeding the arrays with Barker codes, as suggested by previous investigations. The present work has direct applications in special public address systems and underwater acoustical arrays.