Modified Learning Vector Quantization Research Articles

Preconverged learning vector quantization detection method for integrated optical fiber-visible light access network

Recently, visible light communication (VLC) has become a hot topic owing to its advantages, such as high security, high capacity, and antielectromagnetic interference. VLC has become a potential candidate for indoor access networks due to the feasibility of combining illumination and communication. However, intersymbol interference and random phase rotation induced by the transmission channel will decrease the system performance. Traditional constant modulus algorithm (CMA) performs well in converging constellation, but random phase rotation still exists since CMA is insensitive to phase. We demonstrate an optical fiber and visible light integrated communication system. 960-Mbps multiband geometrically shaping (GS) quadrature amplitude modulation (QAM)-8 transmission is realized. Modified learning vector quantization is employed after CMA equalization to mitigate the impairment induced by both optical fiber and visible light channel. Experimental results indicate the feasibility of the proposed method. The Q factor of GS QAM-8 could be enhanced by 5.98 dB with the proposed method.

Simulation and Experimental Investigation on Active Thermography Test of the Solder Balls

The miniaturized electronic products require not only high density, but also high reliability. Defect inspection for the high-density package becomes challenging gradually. In this article, active thermography technology is used to detect the solder balls in flip-chip (FC) packages. A single-layer FC model is constructed and heat conduction in the FC package is simulated by using the finite-element analysis code of COMSOL. Experimental investigation is also carried out to inspect the solder defects by using the laser excitation thermography test system. Infrared images of the SFA1 package are captured, and the spatial adaptive filtering algorithm is used to remove the thermal noise. The hot spots of the solder balls are segmented from the filtered infrared image, and the representative features are extracted for each hot spot. The modified learning vector quantization (LVQ) neural network is used for the classification of the solder balls. The missing solder balls are identified accurately. The simulation and experimental results prove that the proposed approach using active thermography and LVQ algorithm is effective for defect inspection in high-density electronic devices.

A Modified Parallel Learning Vector Quantization Algorithm for Real-Time Hardware Applications

In this study a modified learning vector quantization (LVQ) algorithm is proposed. For this purpose, relevance LVQ (RLVQ) algorithm is efficiently combined with a reinforcement mechanism. In this mechanism, it is shown that the proposed algorithm is not affected constantly by both relevance–irrelevance input dimensions and the winning of the same neuron. Hardware design of the proposed scheme is also given to illustrate the performance of the algorithm. The proposed algorithm is compared to the corresponding ones with regard to success rate and running time.

Pertinent Prosodic Features for Speaker Identification by Voice

Most existing systems of speaker recognition use “state of the art” acoustic features. However, many times one can only recognize a speaker by his or her prosodic features, especially by the accent. For this reason, the authors investigate some pertinent prosodic features that can be associated with other classic acoustic features, in order to improve the recognition accuracy. The authors have developed a new prosodic model using a modified LVQ (Learning Vector Quantization) algorithm, which is called MLVQ (Modified LVQ). This model is composed of three reduced prosodic features: the mean of the pitch, original duration, and low-frequency energy. Since these features are heterogeneous, a new optimized metric has been proposed that is called Optimized Distance for Heterogeneous Features (ODHEF). Tests of speaker identification are done on Arabic corpus because the NIST evaluations showed that speaker verification scores depend on the spoken language and that some of the worst scores were got for the Arabic language. Experimental results show good performances of the new prosodic approach.

DCT-Yager FNN: A Novel Yager-Based Fuzzy Neural Network With the Discrete Clustering Technique

Earlier clustering techniques such as the modified learning vector quantization (MLVQ) and the fuzzy Kohonen partitioning (FKP) techniques have focused on the derivation of a certain set of parameters so as to define the fuzzy sets in terms of an algebraic function. The fuzzy membership functions thus generated are uniform, normal, and convex. Since any irregular training data is clustered into uniform fuzzy sets (Gaussian, triangular, or trapezoidal), the clustering may not be exact and some amount of information may be lost. In this paper, two clustering techniques using a Kohonen-like self-organizing neural network architecture, namely, the unsupervised discrete clustering technique (UDCT) and the supervised discrete clustering technique (SDCT), are proposed. The UDCT and SDCT algorithms reduce this data loss by introducing nonuniform, normal fuzzy sets that are not necessarily convex. The training data range is divided into discrete points at equal intervals, and the membership value corresponding to each discrete point is generated. Hence, the fuzzy sets obtained contain pairs of values, each pair corresponding to a discrete point and its membership grade. Thus, it can be argued that fuzzy membership functions generated using this kind of a discrete methodology provide a more accurate representation of the actual input data. This fact has been demonstrated by comparing the membership functions generated by the UDCT and SDCT algorithms against those generated by the MLVQ, FKP, and pseudofuzzy Kohonen partitioning (PFKP) algorithms. In addition to these clustering techniques, a novel pattern classifying network called the Yager fuzzy neural network (FNN) is proposed in this paper. This network corresponds completely to the Yager inference rule and exhibits remarkable generalization abilities. A modified version of the pseudo-outer product (POP)-Yager FNN called the modified Yager FNN is introduced that eliminates the drawbacks of the earlier network and yi- elds superior performance. Extensive experiments have been conducted to test the effectiveness of these two networks, using various clustering algorithms. It follows that the SDCT and UDCT clustering algorithms are particularly suited to networks based on the Yager inference rule.

POP-Yager: A novel self-organizing fuzzy neural network based on the Yager inference

A Pseudo-Outer Product based Fuzzy neural network using the Yager Rule of Inference [Keller, J. M., Yager, R. R., & Tahani, H. (1992). Neural network implementation of fuzzy logic. Fuzzy Sets and Systems, 45(1), 1-12.] called the POP-Yager FNN is proposed in this paper. The proposed POP-Yager FNN training consists of two phases; namely: the fuzzy membership derivation phase using the Modified Learning Vector Quantization (MLVQ) method [Ang, K. K. (1998). POPFNN-CRI(S): A Fuzzy neural network based on the compositional rule of inference. M. Phil. Dissertation, Nanyang Technological University.]; and the rule identification phase using the novel one-pass LazyPOP learning algorithm [Quek, C. & Zhou, R. W. (1999). The POP learning algorithms: Reducing work in identifying fuzzy rules. Neural Networks, 14(10), 1431-1445]. The proposed two-phase learning process effectively constructs the membership functions and identifies the fuzzy rules. Inference process in POP-Yager is based on the well-established Yager fuzzy inference rule [Keller, J. M., Yager, R. R. & Tahani, H. (1992). Neural network implementation of fuzzy logic. Fuzzy Sets and Systems, 45(1), 1-12]. Operations in POP-Yager strictly perform the logical processes of the Yager inference rule. This gives the novel network a strong theoretical foundation. Extensive experimental results based on the classification performance of the POP-Yager FNN using the Anderson's Iris data [Duda, R. O. & Hart, P. E. (1973). Pattern classification and scene analysis. Wiley.] are presented for discussion. Results show that the POP-Yager FNN possesses excellent recall and generalization abilities. The performance of the POP-Yager FNN as a general approximation of traffic flow data [Tan, G. K. (1997). Feasibility of predicting congestion states with neural networks. Final Year Project Report, Nanyang Technological University, CSE.] is analysed.

Behavior Learning of Autonomous Robots by Modified Learning Vector Quantization

Behavior Learning of Autonomous Robots by Modified Learning Vector Quantization

Damage characterization of carbon/carbon laminates using neural network techniques on AE signals

Carbon/Carbon (C/C) composites have a large number of microcracks in matrix-matrix or fiber-matrix regions resulting from the thermal processes during manufacturing. Although not harmful to the overall structural integrity, such a network of microcracks creates a noisy background from early load application which ‘covers’ the onset of critical failure mechanisms. Conventional AE analysis based on a sudden activity increase observed in the cumulative events versus load plot, or amplitude distribution histograms, provides limited information on this type of AE activity. Instead, multivariate techniques of unsupervised pattern recognition, taking into account a large number of AE signal descriptors, are proven useful for the clustering of similar AE events. AE results from a systematic fracture mechanics study of 2D woven C/C laminates are analyzed in this paper. Artificial Neural System (ANS) methods are employed for the clustering of similar AE signals, enabling a phenomenological correlation with the actual failure modes. The numerical procedure introduces a modified Learning Vector Quantization (LVQ) technique which was proved fast and suitable for the type of AE data emitted by composites. Fractographic evidence from failed tensile coupons corroborates the predictions of the numerical method in recognizing different failure mechanisms. Cumulative event charts of the various classes versus load demonstrate the criticality of each class on final coupon failure, and may lead to the definition of reliable criteria for the evaluation of remaining strength or life.

An Application of Wavelet-Based Vector Quantization in Target Recognition

An automatic target recognition (ATR) classifier is constructed that uses a set of dedicated vector quantizers (VQs). The background pixels in each input image are properly clipped out by a set of aspect windows. The extracted target area for each aspect window is then enlarged to a fixed size, after which a wavelet decomposition splits the enlarged extraction into several subbands. A dedicated VQ codebook is generated for each subband of a particular target class at a specific range of aspects. Thus, each codebook consists of a set of feature templates that are iteratively adapted to represent a particular subband of a given target class at a specific range of aspects. These templates are then further trained by a modified learning vector quantization (LVQ) algorithm that enhances their discriminatory characteristics.

A CONTINUOUS SPEECH RECOGNITION SYSTEM USING A MODIFIED LVQ2 METHOD AND A DEPENDENCY GRAMMAR WITH SEMANTIC CONSTRAINTS

This paper describes an overview of a continuous speech recognition system composed of an acoustic processor and a linguistic processor. The system deals with 843 conceptual words and 431 functional words. We have constructed an acoustic processor using a modified learning vector quantization method (MLVQ2) for phoneme recognition. The phoneme recognition score was 85.5% for 226 sentences uttered by two male speakers. The linguistic processor is composed of a processor for spotting bunsetsu units (i.e. units similar to a “phrase” in English) and a syntactic processor. The structure of the bunsetsu unit is effectively described by a finite-state automaton, the test-set word-perplexity of which is 230. In the processor for spotting bunsetsu units, using a syntax-driven continuous-DP matching algorithm, the bunsetsu units are spotted from a recognized phoneme sequence and then a bunsetsu unit lattice is generated. In the syntactic processor, the bunsetsu unit lattice is parsed based on the dependency grammar, which is expressed as the correspondence between a FEATURE marker in a modifier-bunsetsu and a SLOT-FILLER marker in a head-bunsetsu. The recognition scores of the bunsetsu units and conceptual words were 75.2% and 88.9% respectively for 226 sentences uttered by the two male speakers.