Udoinyang G. Inyang

A hybrid knowledge discovery system for oil spillage risks pattern classification

The complexity and the dynamism of oil spillages make it difficult for planners and responders to produce robust plans towardstheir management. There is need for an understanding of the nature, sources, impact and responses required to prevent or controltheir occurrence. This paper develops an intelligent hybrid system driven by Sugeno-Type Adaptive Neuro Fuzzy InferenceSystem (ANFIS) for the identification, extraction and classification of oil spillage risk patterns. Dataset consisting of 1008records was used for training, validation and testing of the system. Result of sensitivity analysis shows that Cause, Locationand Type of spilled oil have cumulative significance of 85.1%. Optimal weights of Neural Network (NN) were determined viaGenetic Algorithm with hybrid encoding scheme. The Mean Squared Error (MSE) of NN training is 0.2405. NN training,validation and testing results yielded R > 0.839 in all cases indicating a strong linear relationship between each output andtarget data. Rule pruning was performed with support (15%) and confidence (10%) minimum thresholds and antecedent-size of3. The performance of the ANFIS was evaluated with eight different types of membership functions (MFs) and two learningalgorithms. The model with triangular MF gave the best performance among all other given models while hybrid-learningalgorithm performed better than back propagation algorithm. The ANFIS model reported in the paper adopted triangular MFand hybrid learning algorithm for the predication and classification of oil spillage risk patterns. Average training and testingMSE of the model is 0.414315 and 0.221402 respectively. The knowledge mining results show that ANFIS based systemsprovide satisfactory results in the prediction and classification of oil spillage risk patterns.

Unsupervised mining of under-resourced speech corpora for tone features classification

In this contribution, the unsupervised mining of speech corpora for the efficient classification of tone features was investigated. Input vectors to the experiment were generated from tone pattern alignments of Ibibio (Benue-Congo, Nigeria) corpus. The corpus used for the experiment contained 16,905 words/phrases. The proposed system design is novel, and integrates two unsupervised tools - k-means clustering and self organizing map (SOM) model, into a methodological workflow, that evaluates and selects the optimal number of clusters with the subsequent association of each clustering point to the input data points. In order to reduce data dimensionality for effective visualization, a non-negative matrix factorization (NMF) was introduced to rid the k-means clusters of noisy attributes. The k-means cluster points generated by the optimum clusters (two in this case) were evaluated by the Silhouette algorithm and finally fed into the SOM, to improve the efficiency of features classification. Results obtained validate existing research claims and demonstrates the importance of vowel-only features in the recognition of tone patterns. A SOM visualization of the input vectors revealed that vowel-only feature correlates better with other input vectors such as syllable and phoneme, compared to consonant-only features. Furthermore, clustering the input datasets into the optimal number of clusters enabled proper and timely visualization of the map. This contribution is therefore vital for advancing future speech processing research on under-resourced languages.

Towards a Hybrid Learning Approach to Efficient Tone Pattern Recognition

Tone has remained an interesting puzzle to the development of language resources for African languages, mainly because its appearance (within a word) is not segmentally fixed. In this contribution, we begin by proposing a tone marking framework that intelligently tags an input corpus using a close-copy synthesis of tone-tags generated by a Hidden Markov Model (HMM) syllabifier. Next, we investigate the recognition of tone patterns by building a generic architecture that will serve diverse languages. The proposed architecture is a multi-layer feedforward neural network implementing the Levenberg-Marquardt backpropagation algorithm. The network consists of, (i) seventeen inputs describing the tone patterns of Ibibio (ISO 693-3: nic; Ethnologue: IBB), with training data captured from an input corpus of 16,905 phrases; (ii) a target class that learns tone recognition from a combination of the input tone patterns and boundary tone – an important feature used for intonation analysis. Results obtained showed that our tone marking model perfectly tagged the input corpus, except for phonemes with more than one diacritic marks. Concerning the recognition of tone patterns, we deduced from a confusion matrix that 93.1 % of the tone patterns were correctly classified, while the remaining 6.9 % of the patterns were misclassified. A greater chunk of the misclassified cases came from non-boundary tone information, which presence inhibits speech quality. The ROC curve also showed good classification of the training, testing and validation datasets. A future direction of this paper is the introduction of an unsupervised solution and additional tone-bearing information such as syllables and vowels, to improve the learning system; and a comparison of our approach with other methods.

Adaptive Prosody Modelling for Improved Synthetic Speech Quality

Neural networks and fuzzy logic have proven to be efficient when applied individually to a variety of domain-specific problems, but their precision is enhanced when hybridized. This contribution presents a combined framework for improving the accuracy of prosodic models. It adopts the Adaptive Neuro-fuzzy Inference System (ANFIS), to offer self-tuned cognitive-learning capabilities, suitable for predicting the imprecise nature of speech prosody. After initializing the Fuzzy Inference System (FIS) structure, an Ibibio (ISO 693–3: nic; Ethnologue: IBB) speech dataset was trained using the gradient descent and non-negative least squares estimator (LSE) to demonstrate the feasibility of the proposed model. The model was then validated using synthesized speech corpus dataset of fundamental frequency (F0) values of ibibio tones, captured at various contour positions (initial, mid, final) within the courpus. Results obtained showed an insignificant difference between the predicted output and the check dataset with a checking error of 0.0412, and validates our claim that the proposed model is satisfactory and suitable for improving prosody prediction of synthetic speech.

Experimental study of neuro-fuzzy-genetic framework for oil spillage risk management

This paper reports the findings from the experimental study of an intelligent system driven by Neural Network (NN), Fuzzy Logic (FL) and Genetic Algorithm (GA) for knowledge discovery and oil spillage risk management. Application software was developed in an environment characterized by 11Ants Analytics, Matrix Laboratory (MatLab), Microsoft Excel, SPSS and GraphPadInstat as frontend engines; Microsoft Access Database Management System as backend engine and Microsoft Windows as platform. 11Ants Analytics served as a tool for oil spillage indicators rank analysis and predictive model building. Matlab served as a tool for the extraction of patterns from 11Ants Analytics Model of oil spillage. Microsoft Excel serves as an interface between 11Ants Analytics and Matlab. Microsoft Excel, SPSS and GraphPadInstat serve as tools for the generation of relevant statistics. Indicators of oil spillage risks serve as input to the NN. GA is used to provide optimal set of parameters for NN training while FL used for modelling imprecise knowledge and provision of membership functions for the GA and NN. Data on Oil Spill incidences associated with oil exploration activities in the Niger Delta Region of Nigeria were collected from National Oil Spill Detection and Response Agency (NOSDRA) and used to assess and evaluate the practical function of the intelligent system. Adaptive Neuro-Fuzzy Inference System (ANFIS) driven by Mamdani’s inference mechanism was used to predict and estimate oil spillage risks. The findings from the experimental study are presented.

Unsupervised visualization of Under-resourced speech prosody

Abstract In this paper, an unsupervised visualization framework for analyzing under-resourced speech prosody is proposed. An experiment was carried out for Ibibio–a Lower Cross Language of the New Benue Congo family, spoken in the Southeast coastal region of Nigeria, West Africa. The proposed methodology adopts machine learning, with semi-automated procedure for extracting prosodic features from a translated prosodically stable corpus ‘The Tiger and the Mouse’—a text corpus that demonstrates the prosody of read-aloud English. A self-organizing map (SOM) was used to learn the classification of certain input vectors (speech duration, fundamental frequency: F0, phoneme pattern (vowels only), tone pattern), and provide visualization of the clusters structure. Results obtained from the experiment showed that duration and F0 features realized from speech syllables are indispensable for modeling phoneme and tone patterns, but the tone input classes revealed clusters with well separated boundaries and well distributed component planes, compared to the phoneme input classes. Further, except for very few outliers, the map weights were well distributed with proper neighboring neuron connections across the input space. A possible future work to advance this research is the development of the languages corpus, for the discovery of prosodic patterns in expressive speech.

Speech Features Analysis for Tone Language Speaker Discrimination Systems

In this paper, a speech pattern analysis framework for tone language speaker discrimination systems is proposed. We hold the hypothesis that speech feature variability is an efficient means for discriminating speakers. To achieve this, we exploit prosody-related acoustic features (pitch, intensity and glottal pulse) of corpus recordings obtained from male and female speakers of varying age categories: children (0–15), youths (16–30), adults (31–50), seniors (above 50)—and captured under suboptimal conditions. The speaker dataset was segmented into three sets: train, validation and test set—in the ratio of 70%, 15% and 15%, respectively. A 41 × 14 self-organizing map (SOM) architecture was then used to model the speech features, thereby determining the relationship between the speech features, segments and patterns. Results of a speech pattern analysis indicated wide F0 variability amongst children speakers compared with other speakers. This gap however closes as the speaker ages. Further, the intensity variability among speakers was similar across all speaker classes/categories, while glottal pulse exhibited significant variation among the different speaker classes. Results of SOM feature visualization confirmed high inter-variability—between speakers, and low intra-variability—within speakers.

Intra-Speaker Variability Assessment for Speaker Recognition in Degraded Conditions: A Case of African Tone Languages

An extensive study on intra-speaker variability is presented in this chapter. The chapter is organized as follows: Section 2.1 gives a succinct background on speaker recognition and recent applications. Section 2.2 reviews related works on speaker perception, channel variability, as well as clustering and visualization. Section 2.3 provides an in-depth discussion on the phonology (study of sound patterns and their meanings) of Ibibio – a low-resourced language – spoken in the Southeast coastal region of Nigeria, West Africa – and used in this study to demonstrate intra-speaker variability. Section 2.4 is the methodology of the study and includes: the system’s architecture, utterance dataset formation, participants/speakers selection – speech recording – speech signal dataset processing, speech feature extraction, and speech feature dataset construction. Section 2.5 presents the results obtained from (i) a frame-by-frame analysis, (ii) principal component analysis (PCA), and (iii) self-organizing map (SOM) clustering and visualization – on the extracted speech features (duration, F0, intensity, formants, pulses and MFCC: mel-cepstral coefficient). The chapter ends with a conclusion and pointer to future research direction.

Comparative Analysis of Neural Network Models for Petroleum Products Pipeline Monitoring

In recent years, Neural Network (NN) has gained popularity in proffering solution to complex nonlinear problems. Monitoring of variations in Petroleum Products Pipeline (PPP) attributes (flow rate, pressure, temperature, viscosity, density, inlet and outlet volume) which changes with time is complex due to existence of non linear interaction amongst the attributes. The existing works on PPP monitoring are limited by lack of capabilities for pattern recognition and learning from previous data. In this paper, NN models with pattern recognition and learning capabilities are compared with a view of selecting the best model for monitoring PPP. Data was collected from Pipelines and Products Marketing Company (PPMC), Port Harcourt, Nigeria. The data was used for NN training, validation and testing with different NN models such as Multilayer Perceptron (MLP), Radial Basis Function (RBF), Generalized Feed Forward (GFF), Support Vector Machine (SVM), Time Delay Network (TDN) and Recurrent Neural Network (RNN). Neuro Solutions 6.0 was used as the front-end-engine for NN training, validation and testing while My Structured Query Language (MySQL) database served as the back-end-engine. Performance of NN models was measured using Mean Squared Error (MSE), Mean Absolute Error (MAE), Correlation Coefficient ( r ), Akaike Information Criteria (AIC) and Minimum Descriptive Length (MDL). MLP with one hidden layer and three processing elements performed better than other NN models in terms of MSE, MAE, AIC, MDL and r values between the computed and the desired output.

Intelligent Speech Features Mining for Robust Synthesis System Evaluation.

Speech synthesis evaluation involves the analytical description of useful features, sufficient to assess the performance of a speech synthesis system. Its primary focus is to determine the degree of semblance of synthetic voice to a natural or human voice. The task of evaluation is usually driven by two methods: the subjective and objective methods, which have indeed become a regular standard for evaluating voice quality, but are mostly challenged by high speech variability as well as human discernment errors. Machine learning (ML) techniques have proven to be successful in the determination and enhancement of speech quality. Hence, this contribution utilizes both supervised and unsupervised ML tools to recognize and classify speech quality classes. Data were collected from a listening test (experiment) and the speech quality assessed by domain experts for naturalness, intelligibility, comprehensibility, as well as, tone, vowel and consonant correctness. During the pre-processing stage, a Principal Component Analysis (PCA) identified 4 principal components (intelligibility, naturalness, comprehensibility and tone) – accounting for 76.79% variability in the dataset. An unsupervised visualization using self organizing map (SOM), then discovered five distinct target clusters with high densities of instances, and showed modest correlation between significant input factors. A Pattern recognition using deep neural network (DNN), produced a confusion matrix with an overall performance accuracy of 93.1%, thus signifying an excellent classification system.