Recent Trends in Intelligent and Emerging Systems

Since stochastic wireless channels are highly random, fuzzy-based system are suitable options to deal with such uncertainty. This is because of the fact that the fuzzy system provides expert-level decision while tracking microscopic changes. Fuzzy system, however, requires support from artificial neural network (ANN)s for implementing inference rules. When fuzzy and ANN systems are combined, either neuro-fuzzy (NF) or fuzzy-neural (FN) frameworks are derived. Here, we propose an NF-based model for data recovery in multi-antenna setups when transmitted through stochastic wireless channels. Experimental results show that the proposed approach is computationally efficient.

Diversity techniques can be used to reduce the ill effects of multipath fading observed in wireless channels. Multiple-input-multiple-output (MIMO) technology is a promising application of multiple antennas at both transmitter and receiver to improve communication performance by achieving spatial diversity. The concept of space–time coding has arisen from diversity techniques using smart antennas. With the implementation of data coding and signal processing at both sides of transmitter and receiver, space–time coding now is more effective than traditional diversity techniques. Space–time block codes (STBC) were designed to achieve the maximum diversity order for the given number of transmit and receive antennas subject to the constraint of having a simple linear decoding algorithm. Application of ANNs for STBC decoding is such an area which offers solutions to tackle the intricacies associated with the fluctuations observed in multipath propagation which is always the problem area in wireless communication.

In this chapter, the design of a digital receiver for carrier phase tracking is presented. The receiver architecture includes a least square polynomial fitting (LSPF)-based digital phase-locked loop (DPLL). Bit error rate (BER) performance of the proposed system for dealing with Rayleigh and Rician fading for different numbers of paths with coded and uncoded channel is presented here. The performance of the DPLL for carrier phase tracking with signal using QPSK modulation transmitted through Rayleigh and Rician fading channels are compared with coded and uncoded conditions. Simulation results show that the proposed DPLL-based approach shows significant improvement using BCH coding both in Rayleigh and Rician fading channels. Several essential processes like noise and CCI cancelation, equalization, etc., that are integral to the traditional frameworks are made redundant by the proposed DPLL-based approach. The composite outcome of these separate processes is combined by the DPLL action making it a reliable and efficient mechanism leading to a compact design.

This work is related to design certain adaptive equalization and error correction coding and aids to maximal-ratio combining (MRC) in faded wireless channels. The performances derived are analyzed using semianalytic (SA) and Monte Carlo (MC) approaches in order to achieve improvement in bit error rates (BERs) of demodulated signals in wireless channels that have both Gaussian and multipath fading characteristics. Modulation techniques used in this work are bipolar phase shift keying (BPSK), quadrature phase shift keying (QPSK), and differential phase shift keying (DPSK). The work though considers the use of least mean square (LMS), adaptive filter blocks as part of a MRC setup and is tested under SNR variation between \(-\)10 and 10 dB. The results generated justify the use of the adaptive equalizer block as an aid to the MRC setup. The validity of the results is further confirmed by comparing to those obtained via SA approach and MC simulations.

Ultra wide band (UWB) communication systems occupy huge bandwidths with very low power spectral densities. In a high data rate UWB indoor communication system, the delay spread due to multipath propagation results in inter symbol interference (ISI) which can significantly increase the bit error rate (BER). The distortion and fading caused by the UWB channel and noise sources is removed by equalization which is a signal processing technique. In this work, IEEE 802.15.4a UWB channel model is used for both LOS and NLOS residential environment in a frequency range 2–10 GHz. QAM modulation is used to transmit large volume of data at a time. Equalization is carried out using zero forcing (ZF) and minimum mean square error estimation (MMSE) algorithms. MMSE algorithm shows better performance than the ZF algorithms by reducing BER.

Birth-cry, birth-weight, mother’s distress during pregnancy, baby’s health condition soon after birth, are some symptoms that might have some relationship with disability in a child. Influence factors are determined and multiple linear regression and backpropagation artificial neural network (ANN) are applied for modeling the occurrence of disability in a child. Results of multiple regression show that the factors considered have significant effects on the occurrence of disability. Also, the largest beta value (regression coefficient) corresponds to the birth-cry factor of a newborn. It implies the strongest and unique contribution of this variable to explain the dependent variable, which in this case is the proportion of disabled children. An ANN in feedforward form is also configured to perform identical regression for the purpose. Experimental results show that the ANN is a suitable technique for the study of such cases.

Alzheimer’s disease is a common disease which is characterized by a person losing his memory progressively. Finally, the person also loses his life. It is often seen in the people above the age 60 but it may occur early. This disease destroys memory cells of the brain. Till now, it is a disease without any treatment and also there are no proper means of diagnosis. Research shows that most often it occurs either due to the deposition of defective structure of amyloid protein or due to the tangles in the brain. In this paper, we have proposed a system to detect the defective Amyloid protein using two classifiers. Secondary structure of Amyloid protein is detected and analyzed in our work which provides a way to predict the cause of Alzheimer.

Very often, industrial sickness is identified using certain traditional techniques which rely upon a range of manual monitoring and compilation of financial records. It makes the process tedious, time consuming, and often are susceptible to manipulation. Hence, decision makers, planners, and funding agencies of such units are sometimes surrounded by uncertainty and unpredictable situations while taking decisions regarding the state of industrial health and the subsequent measures required. Therefore, certain readily available tools are required which can deal with such uncertain situations arising out of industrial sickness. It is more significant for a country like India where the fruits of developments are rarely equally distributed. In this paper, we propose an approach based on certain soft-computational tools specially using Artificial neural network (ANN) to deal with industrial sickness with specific focus on a few such units taken from a less-developed northeast (NE) Indian state like Assam. More specifically, we here propose, a soft-computational tool which formulates certain decision support mechanism to decide upon industrial sickness using eight different parameters which are directly related to the stages of sickness of such units. The mechanism primarily identifies a few stages of industrial health using various inputs provided in terms of the eight identified parameters. This decision is further compared with the results obtained from another set of ANNs where the model uses certain signals and symptoms of industrial health to decide upon the state of a unit. Specifically, we train multiple ANN blocks with data obtained from a few selected units of Assam so that required decisions related to industrial health could be taken. The system thus formulated could become an important part of planning and development. It can also contribute toward computerization of decision support systems related to industrial health and help in better management.

Hand gesture recognition system is an essential element used for human–computer interaction (HCI). The use of hand gestures provides an attractive alternative to cumbersome interface devices for HCI. Proper hand segmentation from the background and other body parts of a video stream is the primary requirement of the design of a hand gesture-based application. These video frames are captured from a low-cost webcam (camera) for using in a vision-based gesture recognition system. This work reports the design of a continuous hand gesture recognition system. The paper also includes the description of a robust and efficient hand segmentation algorithm where a new method for hand segmentation using different color space models as required by morphological processing are utilized. Problems such as skin color detection, complex background removal, and variable lighting conditions are found to be efficiently handled by this system. Noise present in the segmented image due to dynamic background can be removed with the help of this adaptive technique. The proposed approach is found to be effective for a range of conditions.

Real-time applications like speech processing and image processing are known to have hardware dependencies. Again artificial neural network (ANN)-based recognition systems show dependance on data and hardware for achieving better performance. Therefore, there always exist the possibility of exploring means and methods to evaluate performance difference achieved by ANNs for speech- and image processing applications when hardware framework is varied and at times enhanced and expanded. Formation of certain parallel processing architectures based on multicore layouts are described here. With such frameworks, certain ANN-based processing of speech and image inputs are carried out. The results derived show that the capability of the ANN improves with large sizes of data and expanded hardware layouts. Similar approach is also followed using digital signal processor (DSP). The experimental study shows that compared to conventional CPUs, DSP processor architectures like TMS320C6713 provide greater processing speed and throughput.

In this paper, we present a comparison between a few clustering algorithms including K-means clustering (KMC), Artificial Neural Network (ANN)-based Self-Organization Map (SOM), and Fuzzy C-means (FCM) clustering for the determination of number of phonemes present in a spoken Assamese word. Here, a block is designed to determine the number of phonemes present in a particular speech dataset. The phoneme count determination technique takes some initial decision about the possible number of phonemes present in a particular word. Comparing the success rate of correct decision from the proposed clustering techniques it is observed that the SOM-based technique provides more correct decisions compared to KMC-based technique and FCM-based approach provides even better decisions than the SOM-based technique. The results show that FCM generates better performance for all the cases considered.

Empirical mode decomposition (EMD) is a method for nonstationary signal analysis, where signals are decomposed into number of high-frequency modes called intrinsic mode function (IMF)s and a low-frequency component called the residual. If this residual is considered as the source signal in case of a speech signal and vocal tract filter response is estimated, the original signal can be reconstructed. The nonstationary attribute of a speech signal restricts direct application of the conventional digital signal processing (DSP) techniques to a speech signal. Since EMD performs decomposition assuming the nonstationary nature of a signal, it has been observed that frame-by-frame analysis is not required in the proposed reconstruction model. The effectiveness of reconstruction using EMD residual is experimented in case of speech data collected in various conditions like clean, noisy, mobile channel including speaker’s mood variation.

This work focuses on the classification of Assamese vowel speech and recognition using Gaussian mixture model (GMM). The results are compared to the results obtained using artificial neural network (ANN). The training data is composed of a database of eight different vowels of Assamese language with 10 different recorded speech samples of each vowel as a set in noise-free and noisy environments. The testing data similarly is composed of the same number of vowels with each vowel containing 23 different recorded samples. Cepstral mean normalization (CMN) and maximum likelihood linear regression (MLLR) are used for speech enhancement of the data which is degraded due to noise. Feature extraction is done using mel frequency cepstral coefficients (MFCC). GMM and ANN approaches are used as classifiers for an automatic speech recognition (ASR) system. We found the success rate of the GMM to be around 81 \(\%\) and that of the ANN to be above 85 \(\%\).

In this paper, a brief overview derived out of detailed survey of speech recognition works reported in Indian languages is described. Robustness of speech recognition systems toward language variation is the recent trend of research in speech recognition technology. To develop a system which can communicate with human in any language like any other human is the foremost requirement in order to design appropriate speech recognition technology for one to all. India is a country which has vast linguistic variations among its billion plus population. Therefore, it provides a sound area of research toward language-specific speech recognition technology. From the beginning of the commercial availability of the speech recognition system, the technology has been dominated by the hidden Markov model (HMM) methodology due to its capability of modeling temporal structures of speech and encoding them as a sequence of spectral vectors. Most of the work done in Indian languages also uses HMM technology. However, from the last 10–15 years after the acceptance of neurocomputing as an alternative to HMM, artificial neural network (ANN)-based methodologies have started to receive attention for application in speech recognition. This is a trend worldwide as part of which few works have also been reported by a few researchers.

Power-aware or power-conscious very-large-scale integration (VLSI) design is gradually evolving to be one of the most pronounced and important, yet challenging aspect of system design and implementation. The issue of power is being addressed by different researchers around the globe at different levels of implementation and fabrication. This chapter highlights the recent works in circulation in open literature primarily reported during the present decade. It focuses broadly on four important aspects viz. device design, circuit design, high data rate applications, and synthesis and scheduling. The objective is to discuss the relevant developments as well as to identify existing challenges faced by the community of researchers engaged in power-aware VLSI design.

The proliferation of number of users in a limited wireless spectrum have raised the levels of inter symbol interference (ISI) and have also contributed towards probable degradation of quality of service (QoS). The key challenges faced by upcoming wireless communication systems is to provide high-data-rate wireless access with better QoS. Also, the fast shrinking spectrum for such communication have necessitated the development of methods to increase spectral efficiency. Multiple input multiple output (MIMO) wireless technology is a viable option in such a situation and is likely to be able to meet the demands of these ever-expanding mobile networks. Many researchers have explored this field over a considerable period of time. A sizable portion of the research have been on the application of traditional statistical methods in such areas. Over the years, soft computational tools like artificial neural network (ANN), fuzzy systems and their combinations have received attention in the diverse segments of wireless communication. This is because of the fact that these are learning based systems. These learn from the environment, retain the knowledge and use it subsequently. This paper highlights some of the important application areas in wireless communication which have reported the use of soft computing tools in wireless communication that are in circulation in open literature.