Phonetic segmentation using multiple speech features

In this paper we propose a method for improving the performance of the segmentation of speech waveforms to phonetic units. The proposed method is based on the well known Viterbi time-alignment algorithm and utilizes the phonetic boundary predictions from multiple speech parameterization techniques. Specifically, we utilize the most appropriate, with respect to boundary type, phone transition position prediction as initial point to start Viterbi time-alignment for the prediction of the successor phonetic boundary. The proposed method was evaluated on the TIMIT database, with the exploitation of several, well known in the area of speech processing, Fourier-based and wavelet-based speech parameterization algorithms. The experimental results for the tolerance of 20 milliseconds indicated an improvement of the absolute segmentation accuracy of approximately 0.70%, when compared to the baseline speech segmentation scheme.     

Text-independent speaker identification using Radon and discrete cosine transforms based features from speech spectrogram

This paper presents a new feature extraction technique for speaker recognition using Radon transform (RT) and discrete cosine transform (DCT). The spectrogram is compact, efficient in representation and carries information about acoustic features in the form of pattern. In the proposed method, speaker specific features have been extracted by applying image processing techniques to the pattern available in the spectrogram. Radon transform has been used to derive the effective acoustic features from the speech spectrogram. Radon transform adds up the pixel values in the given image along a straight line in a particular direction and at a specific displacement. The proposed technique computes Radon projections for seven orientations and captures the acoustic characteristics of the spectrogram. DCT applied on Radon projections yields low dimensional feature vector. The technique is computationally efficient, text-independent, robust to session variations and insensitive to additive noise. The performance of the proposed algorithm has been evaluated using the Texas Instruments and Massachusetts Institute of Technology (TIMIT) and our own created Shri Guru Gobind Singhji (SGGS) databases. The recognition rate of the proposed algorithm on TIMIT database (consisting of 630 speakers) is 96.69% and for SGGS database (consisting of 151 speakers) is 98.41%. These results highlight the superiority of the proposed method over some of the existing algorithms.     

Artificial immune algorithm for IIR filter design

Over the recent years, several studies have been carried out by the researchers to describe a general, flexible and powerful design method based on modern heuristic optimisation algorithms for infinite impulse response (IIR) digital filters since these algorithms have the ability of finding global optimal solution in a nonlinear search space. One of the modern heuristic algorithms is the artificial immune algorithm which implements a learning technique inspired by human immune system. However, the immune system has not attracted the same kind of interest from researchers as other heuristic algorithms. In this work, an artificial immune algorithm is described and applied to the design of IIR filters, and its performance is compared to that of genetic and touring ant colony optimisation algorithms.     

The parameter extraction of the thermally annealed Schottky barrier diode using the modified artificial bee colony

In this paper, a new method based on the modified artificial bee colony (MABC) algorithm to determine the main characteristic parameters of the Schottky barrier diode such as barrier height, ideality factor and series resistance. For this model, the Ni/n-GaAs/In Schottky barrier diode was produced and annealed at different temperature in a laboratory. The performance of the modified ABC method was compared to that of the basic artificial bee colony (ABC), particle swarm optimization (PSO), differential evolution (DE), genetic algorithm (GA) and simulated annealing (SA). From the results, it is concluded that the modified ABC algorithm is more flexible and effective for the parameter determination than the other algorithms.     

Three self-adaptive multi-objective evolutionary algorithms for a triple-objective project scheduling problem

Finding a Pareto-optimal frontier is widely favorable among researchers to model existing conflict objectives in an optimization problem. Project scheduling is a well-known problem in which investigating a combination of goals eventuate in a more real situation. Although there are many different types of objectives based on the situation on hand, three basic objectives are the most common in the literature of the project scheduling problem. These objectives are: (i) the minimization of the makespan, (ii) the minimization of the total cost associated with the resources, and (iii) the minimization of the variability in resources usage. In this paper, three genetic-based algorithms are proposed for approximating the Pareto-optimal frontier in project scheduling problem where the above three objectives are simultaneously considered. For the above problem, three self-adaptive genetic algorithms, namely (i) A two-stage multi-population genetic algorithm (MPGA), (ii) a two-phase subpopulation genetic algorithm (TPSPGA), and (iii) a non-dominated ranked genetic algorithm (NRGA) are developed. The algorithms are tested using a set of instances built from benchmark instances existing in the literature. The performances of the algorithms are evaluated using five performance metrics proposed in the literature. Finally according to the technique for order preference by similarity to ideal solution (TOPSIS) the self-adaptive NRGA gained the highest preference rank, followed by the self-adaptive TPSPGA and MPGA, respectively.     

Speaker-adaptive learning of resonance targets in a hidden trajectory model of speech coarticulation

A novel speaker-adaptive learning algorithm is developed and evaluated for a hidden trajectory model of speech coarticulation and reduction. Central to this model is the process of bi-directional (forward and backward) filtering of the vocal tract resonance (VTR) target sequence. The VTR targets are key parameters of the model that control the hidden VTR’s dynamic behavior and the subsequent acoustic properties (those of the cepstral vector sequence). We describe two techniques for training these target parameters: (1) speaker-independent training that averages out the target variability over all speakers in the training set; and (2) speaker-adaptive training that takes into account the variability in the target values among individual speakers. The adaptive learning is applied also to adjust each unknown test speaker’s target values towards their true values. All the learning algorithms make use of the results of accurate VTR tracking as developed in our earlier work. In this paper, we present details of the learning algorithms and the analysis results comparing speaker-independent and speaker-adaptive learning. We also describe TIMIT phone recognition experiments and results, demonstrating consistent superiority of speaker adaptive learning over speaker-independent one measured by the phonetic recognition performance.     

Global optimization of a neural network-hidden Markov model hybrid

The integration of multilayered and recurrent artificial neural networks (ANNs) with hidden Markov models (HMMs) is addressed. ANNs are suitable for approximating functions that compute new acoustic parameters, whereas HMMs have been proven successful at modeling the temporal structure of the speech signal. In the approach described, the ANN outputs constitute the sequence of observation vectors for the HMM. An algorithm is proposed for global optimization of all the parameters. Results on speaker-independent recognition experiments using this integrated ANN-HMM system on the TIMIT continuous speech database are reported.     

Advanced classification approach for neuronal phoneme recognition system based on efficient constructive training algorithm

This paper introduces a neural network optimization procedure allowing the generation of multilayer perceptron (MLP) network topologies with few connections, low complexity and high classification performance for phoneme’s recognition. An efficient constructive algorithm with incremental training using a new proposed Frame by Frame Neural Networks (FFNN) classification approach for automatic phoneme recognition is thus proposed. It is based on a novel recruiting hidden neuron’s procedure for a single hidden-layer. After an initializing phase started with initial small number of hidden neurons, this algorithm allows the Neural Networks (NNs) to adjust automatically its parameters during the training phase. The modular FFNN classification method is then constructed and tested to recognize 5 broad phonetic classes extracted from the TIMIT database. In order to take into account the speech variability related to the coarticulation effect, a Context Window of Three Successive Frame’s (CWTSF) analysis is applied. Although, an important reduction of the computational training time is observed, this technique penalized the overall Phone Recognition Rate (PRR) and increased the complexity of the recognition system. To alleviate these limitations, two feature dimensionality reduction techniques respectively based on Principal Component Analysis (PCA) and Self Organizing Maps (SOM) are investigated. It is observed an important improvement in the performance of the recognition system when the PCA technique is applied. Optimal neuronal phone recognition architecture is finally derived according to the following criteria: best PRR, minimum computational training time and complexity of the BPNN architecture.     

Phoneme recognition using an adaptive supervised manifold learning algorithm

To effectively handle speech data lying on a nonlinear manifold embedded in a high-dimensional acoustic space, in this paper, an adaptive supervised manifold learning algorithm based on locally linear embedding (LLE) for nonlinear dimensionality reduction is proposed to extract the low-dimensional embedded data representations for phoneme recognition. The proposed method aims to make the interclass dissimilarity maximized, while the intraclass dissimilarity minimized in order to promote the discriminating power and generalization ability of the low-dimensional embedded data representations. The performance of the proposed method is compared with five well-known dimensionality reduction methods, i.e., principal component analysis, linear discriminant analysis, isometric mapping (Isomap), LLE as well as the original supervised LLE. Experimental results on three benchmarking speech databases, i.e., the Deterding database, the DARPA TIMIT database, and the ISOLET E-set database, demonstrate that the proposed method obtains promising performance on the phoneme recognition task, outperforming the other used methods.     

自适应帧长特征提取对语音识别系统性能的影响

提出一种自适应帧长语音特征分析方法,使语音编码更准确,达到提高语音识别性能无愧迁滤帧检测和过渡语音特征表示两方面。采用了两种特征表示方法。基于ＴＩＭＴＴ语音数据包和自定义的汉语语音数据的单词识别实验表明,这两种表示方面有相同的效果,都能在一定程度上提高知识系统的性能,但计算量稍有区别。基于ＴＩＭＩＴ数据的ＤＨＭＭ系统和ＣＨＭＭ系统的错误率分别下降了１１．２１％和９．５８％,基于自定义数据的ＤＨＭＭ     

人工免疫系统超变异模式识别及应用

描述了人工免疫系统(Artificial Immune System,AIS)原理,在人工免疫系统算法的基础上,对免疫系统的超变异特性进行了算法设计,并针对四种简单信号模式对人工免疫系统普通模式识别算法和超变异模式识别算法进行了比较.结果表明:人工免疫系统普通模式识别算法和超变异算法皆可对四种信号模式进行识别,超变异算法可以快速得到最优抗体,且亲和力优于普通人工免疫系统模式识别算法,并给出了人工免疫系统超变异模式识别算法在碳纤维增强复合材料板疏松缺陷超声信号检测中的应用.     

A Cuckoo Search Detector Generation-based Negative Selection Algorithm

The negative selection algorithm (NSA) is an adaptive technique inspired by how the biological immune system discriminates the self from non-self. It asserts itself as one of the most important algorithms of the artificial immune system. A key element of the NSA is its great dependency on the random detectors in monitoring for any abnormalities. However, these detectors have limited performance. Redundant detectors are generated, leading to difficulties for detectors to effectively occupy the non-self space. To alleviate this problem, we propose the nature-inspired metaheuristic cuckoo search (CS), a stochastic global search algorithm, which improves the random generation of detectors in the NSA. Inbuilt characteristics such as mutation, crossover, and selection operators make the CS attain global convergence. With the use of Lévy flight and a distance measure, efficient detectors are produced. Experimental results show that integrating CS into the negative selection algorithm elevated the detection performance of the NSA, with an average increase of 3.52% detection rate on the tested datasets. The proposed method shows superiority over other models, and detection rates of 98% and 99.29% on Fisher’s IRIS and Breast Cancer datasets, respectively. Thus, the generation of highest detection rates and lowest false alarm rates can be achieved.     

基于通用背景-联合估计(UB-JE)的说话人识别方法

在说话人识别中,有效的识别方法是核心.近年来,基于总变化因子分析（i-vector）方法成为了说话人识别领域的主流,其中总变化因子空间的估计是整个算法的关键.本文结合常规的因子分析方法提出一种新的总变化因子空间估计算法,即通用背景—联合估计（Universal background-joint estimation algorithm,UB-JE）算法.首先,根据高斯混合—通用背景模型（Gaussian mixture model-universal background model,GMM-UBM）思想提出总变化矩阵通用背景（UB）算法;其次,根据因子分析理论结合相关文献提出了一种总变化矩阵联合估计（JE）算法;最后,将两种算法相结合得到通用背景—联合估计（UB-JE）算法.采用TIMIT和MDSVC语音数据库,结合i-vector方法将所提的算法与传统算法进行对比实验.结果显示,等错误率（Equal error rate,EER）和最小检测代价函数（Minimum detection cost function,MinDCF）分别提升了8.3%与6.9%,所提方法能够提升i-vector方法的性能.     

An adaptive multi-population genetic algorithm to solve the multi-objective group scheduling problem in hybrid flexible flowshop with sequence-dependent setup times

In this paper we consider a multi-objective group scheduling problem in hybrid flexible flowshop with sequence-dependent setup times by minimizing total weighted tardiness and maximum completion time simultaneously. Whereas these kinds of problems are NP-hard, thus we proposed a multi-population genetic algorithm (MPGA) to search Pareto optimal solution for it. This algorithm comprises two stages. First stage applies combined objective of mentioned objectives and second stage uses previous stage’s results as an initial solution. In the second stage sub-population will be generated by re-arrangement of solutions of first stage. To evaluate performance of the proposed MPGA, it is compared with two distinguished benchmarks, multi-objective genetic algorithm (MOGA) and non-dominated sorting genetic algorithm II (NSGA-II), in three sizes of test problems: small, medium and large. The computational results show that this algorithm performs better than them.     

类搜索算法

提出利用类结构驱动的群体进化计算方法——类搜索算法(CSA).CSA在个体间构造簇类形态的虚拟连接关系,并通过对类组织的结构和类搜索过程进行动态调节来优化模拟进化系统的计算状态,提高群体的搜索效率.介绍了CSA的基本模型,并基于CSA融合进化算子与差分计算机制设计出数值优化算法CSA/DE.对多个典型高纬函数和复杂混合函数的仿真实验结果说明,CSA/DE是一种对高纬连续问题高效、稳定的搜索优化方法.该工作一方面验证了CSA的可行性和有效性;另一方面则显示:基于类搜索模型可有效融合异构且具有不同计算特性的搜索机制,形成对待求解问题更具针对性且协调性更佳的搜索计算方法.这为高性能优化算法的设计提供了一条新的途径.     

A review of clonal selection algorithm and its applications

Recently, clonal selection theory in the immune system has received the attention of researchers and given them inspiration to create algorithms that evolve candidate solutions by means of selection, cloning, and mutation procedures. Moreover, diversity in the population is enabled by means of the receptor editing process. The Clonal Selection Algorithm (CSA) in its canonical form and its various versions are used to solve different types of problems and are reported to perform better compared with other heuristics (i.e., genetic algorithms, neural networks, etc.) in some cases, such as function optimization and pattern recognition. Although the studies related with CSA are increasingly popular, according to our best knowledge, there is no study summarizing the basic features of these algorithms, hybrid algorithms, and the application areas of these algorithms all in one paper. Therefore, this study aims to summarize the powerful characteristics and general review of CSA. In addition, CSA based hybrid algorithms are reviewed, and open research areas are discussed for further research.     

An artificial immune system based algorithm to solve unequal area facility layout problem

This study introduces an artificial immune system (AIS) based algorithm to solve the unequal area facility layout problem (FLP) with flexible bay structure (FBS). The proposed clonal selection algorithm (CSA) has a new encoding and a novel procedure to cope with dummy departments that are introduced to fill the empty space in the facility area. The algorithm showed consistent performance for the 25 test problem cases studied. The problems with 100 and 125 were studied with FBS first time in the literature. CSA provided four new best FBS solutions and reached to sixteen best-so-far FBS solutions. Further, the two very large size test problems were solved first time using FBS representation, and results significantly improved the previous best known solutions. The overall results state that CSA with FBS representation was successful in 95.65% of the test problems when compared with the best-so-far FBS results and 90.90% compared with the best known solutions that have not used FBS representation.     

人工免疫网络记忆分类并行算法研究

人类免疫系统是极为复杂的、固有并行性、分布式系统。人工免疫系统领域已经开发了许多免疫系统启发的算法,但没有几个显示并行性。论文提出并行人工免疫网络记忆分类系统,给出简单的并行人工免疫网络记忆分类算法。初始结果显示,通过简单的并行化方法,与并行人工免疫识别系统AIRS的比较研究表明,并行人工免疫网络记忆分类系统在并行效率等方面的性能优于后者。     

A preference multi-objective optimization based on adaptive rank clone and differential evolution

Evolutionary multi-objective optimization (EMO) algorithms have been used in various real-world applications. However, most of the Pareto domination based multi-objective optimization evolutionary algorithms are not suitable for many-objective optimization. Recently, EMO algorithm incorporated decision maker’s preferences became a new trend for solving many-objective problems and showed a good performance. In this paper, we first use a new selection scheme and an adaptive rank based clone scheme to exploit the dynamic information of the online antibody population. Moreover, a special differential evolution (DE) scheme is combined with directional information by selecting parents for the DE calculation according to the ranks of individuals within a population. So the dominated solutions can learn the information of the non-dominated ones by using directional information. The proposed method has been extensively compared with two-archive algorithm, light beam search non-dominated sorting genetic algorithm II and preference rank immune memory clone selection algorithm over several benchmark multi-objective optimization problems with from two to ten objectives. The experimental results indicate that the proposed algorithm achieves competitive results.     

Application of mutation operators to flower pollination algorithm

Flower pollination algorithm (FPA) is a recent addition to the field of nature inspired computing. The algorithm has been inspired from the pollination process in flowers and has been applied to a large spectra of optimization problems. But it has certain drawbacks which prevents its applications as a standard algorithm. This paper proposes new variants of FPA employing new mutation operators, dynamic switching and improved local search. A comprehensive comparison of proposed algorithms has been done for different population sizes for optimizing seventeen benchmark problems. The best variant among these is adaptive-Lévy flower pollination algorithm (ALFPA) which has been further compared with the well-known algorithms like artificial bee colony (ABC), differential evolution (DE), firefly algorithm (FA), bat algorithm (BA) and grey wolf optimizer (GWO). Numerical results show that ALFPA gives superior performance for standard benchmark functions. The algorithm has also been subjected to statistical tests and again the performance is better than the other algorithms.