Articulatory feature classification using nearest neighbors

An articulatory model of speech production is created for the purpose of studying the links between speech production and perception. A computationally effective method for speech inversion in proposed, using a two-pole predictor structure in order to maintain better articulatory dynamics when compared to conventional dynamic programming methods. Preliminary tests for the effect of inversion are performed for 2500 Finnish syllables extracted from continuous speech, consisting of 125 different syllable classes. A cluster selectivity test shows that the syllables are more reliably clustered using the automatically obtained parametric representation of articulatory gestures rather than the original formant representation that is used as a starting point for the inversion. Index Terms: Articulatory model, speech inversion, motor theory, vocal tract 1. Introduction to articulatory modeling Speech events are more conveniently described in articulatory than acoustic sense. Individual articulators move rather slowly and smoothly when compared to spectral characteristics of speech signals. Since the relative trajectories of different articulators remain rather similar in the production of speech sounds regardless of the speaker, the modeling of speech perception with articulatory modeling may help to overcome many of the problems that arise from the ambiguity in the purely acoustic domain. In the 19th century research on the area of articulatory modeling boomed when the first electrical and then digital models for speech production could be implemented. Researchers have often referred to articulatory models developed by Coker [1], Mermelstein [2] or Maeda [3], for example when studying the speech inverse problem. Maeda’s model’s seven articulatory parameters were estimated from x-ray tracings using so-called arbitrary factor analysis in order to have the parameters maximally uncorrelated to each other. Mermelstein’s geometrical articulatory model depicts the positions of articulators in the midsagittal plane. Lips, jaw, tongue, velum and hyoid are considered as movable structures. In 1990’s and 2000’s more complex vocal tract and tongue models were developed. E.g. Dang and Honda have created a 3D articulatory model which used physiological constraints typical to human articulation in inverting vowel-to-vowel sequences [4].

This paper proposes an approach based on fuzzy rough set theory to improve nearest neighbor based classification. Six measures are introduced to evaluate the quality of the nearest neighbors. This quality is combined with the frequency at which classes occur among the nearest neighbors and the similarity w.r.t. the nearest neighbor, to decide which class to pick among the neighbor's classes. The importance of each aspect is weighted using optimized weights. An experimental study shows that our method, Quality, Frequency and Similarity based Fuzzy Nearest Neighbor (QFSNN), outperforms state-of-the-art nearest neighbor classifiers.

The Nearest Neighbor (NN) classifier is one of the most used and well-known techniques for classification. However, NN suffers from several drawbacks such as high storage requirements, low efficiency in classification response, and noise sensitivity in the neighborhood. Prototype weighting as one the effective learning methods, assigns a weight to each prototype in order to determine its importance and simultaneously removes both redundant and noisy ones as a side effect. NN has an instance-based structure, and this is why; it is one of the rare methods for which the results of cross validation can be achieved for all possible combinations called as Complete-Cross-Validation (CCV) without need to construct and train exponential number of models. Using CCV as the objective function in learning the weights of prototypes can increase generalization capability of the classifier and also decrease the randomness of the learning method. Nevertheless, it cannot be computed such efficiently to be used as a fitness function during most of learning algorithms. In this paper, an iterative learning method has been proposed, in which the objective is incrementally improved without computing CCV even once. The proposed approach was assessed through a series of experiments with UCI in comparison with basic NN and related prototype weighting scheme and could outperform them.

The basic function of the brain is severely affected by alcoholism. For the easy depiction and assessment of the mental condition of a human brain, Electroencephalography (EEG) signals are highly useful as it can record and measure the electrical activities of the brain much to the satisfaction of doctors and researchers. Utilizing the standard conventional techniques is quite hectic to derive the useful information as these signals are highly non-linear and non-stationary in nature. While recording the EEG signals, the activities of the neurons are recorded from various scalp regions which has varied characteristics and has a very low magnitude. Therefore, human interpretation of such signals is very difficult and consumes a lot of time. Hence, with the advent of Computer Aided Diagnosis (CAD) Techniques, identifying the normal versus alcoholic EEG signals has been of great utility in the medical field. In this work, we perform the initial clustering of the alcoholic EEG signals by means of using Correlation Dimension (CD) for easy feature extraction and then the suitable features are selected in it by means of employing various distance metrics like correlation distance, city block distance, cosine distance and chebyshev distance. Proceeding in such a methodology aids and assures that a good discrimination could be achieved between normal and alcoholic EEG signals using non-linear features. Finally, classification is then carried out with the suitable classifiers chosen such as Adaboost.RT classifier, the proposed Modified Adaboost.RT classifier by means of introducing Ridge and Lasso based soft thresholding technique, Random Forest with bootstrap resampling technique, Artificial Neural Networks (ANN) such as Radial Basis Functions (RBF) and Multi-Layer Perceptron (MLP), Support Vector Machine (SVM) with Linear, Polynomial and RBF Kernel, Naïve Bayesian Classifier (NBC), K-means classifier, and K Nearest Neighbor (KNN) Classifier and the results are analyzed. Results report a comparatively high classification accuracy of about 98.99% when correlation distance metrics are utilized with CD and the proposed Modified Adaboost.RT classifier using Ridge based soft thresholding technique.

Machine learning (ML) applications have been thriving recently, largely attributed to the increasing availability of data. However, inconsistency and incomplete information are ubiquitous in real-world datasets, and their impact on ML applications remains elusive. In this paper, we present a formal study of this impact by extending the notion of Certain Answers for Codd tables, which has been explored by the database research community for decades, into the field of machine learning. Specifically, we focus on classification problems and propose the notion of "Certain Predictions" (CP) --- a test data example can be certainly predicted (CP'ed) if all possible classifiers trained on top of all possible worlds induced by the incompleteness of data would yield the same prediction. We study two fundamental CP queries: (Q1) checking query that determines whether a data example can be CP'ed; and (Q2) counting query that computes the number of classifiers that support a particular prediction (i.e., label). Given that general solutions to CP queries are, not surprisingly, hard without assumption over the type of classifier, we further present a case study in the context of nearest neighbor (NN) classifiers, where efficient solutions to CP queries can be developed --- we show that it is possible to answer both queries in linear or polynomial time over exponentially many possible worlds. We demonstrate one example use case of CP in the important application of "data cleaning for machine learning (DC for ML)." We show that our proposed CPClean approach built based on CP can often significantly outperform existing techniques, particularly on datasets with systematic missing values. For example, on 5 datasets with systematic missingness, CPClean (with early termination) closes 100% gap on average by cleaning 36% of dirty data on average, while the best automatic cleaning approach BoostClean can only close 14% gap on average.

Currently, Nearest-Neighbor approaches (NN) have been widely applied to real world image data mining. These approaches have the following three disadvantages: (i) the performance is inferior on small datasets; (ii) the performance of approximated nearest neighbor search will degrade for data with high dimensions; (iii) they are heavily dependent on the chosen feature and distance measure. To overcome these intrinsic weaknesses, we propose a novel Nearest-Neighbor method, which improves the original NN approaches from three aspects. Firstly, we propose a novel neighborhood similarity measure, where the similarity between test images and labeled images in the database is calculated jointly by the original image-to-image similarity and the average similarity of their neighboring unlabeled data. Secondly, we adopt the kernelized locality sensitive hashing to effectively conduct the nearest neighbor search for high dimensional data. Finally, to enhance the robustness of the method on different genres of images, we propose to fuse the discrimination power of different features by considering all the retrieved nearest neighbors via hashing systems using different features/kernels. Experimental result shows the advantage over traditional Nearest-Neighbor methods using the labeled data only. Even when the ratio of labeled data is very small, our method could also achieve remarkable results, thanks to the help of unlabeled data and multiple features.

The nearest neighbor classifier (NNC) is a non-parametric classification technique that classifies a test pattern to the class of its nearest neighbor in the training data. In this research, we applied the NNC to a standard letter recognition data and obtained a superior classification rate in comparison to extant approaches. A prime drawback to the NNC technique has been the relative inefficiency of the model. A modified NNC was implemented and applied to the same recognition problem. It is found that if we choose a suitable threshold minimum difference for classification, we can reduce the CPU time by half without lowering the performance of the classifier.

Combining multiple classifiers is an effective technique for improving accuracy. There are many general combining algorithms, such as Bagging, Boosting, or Error Correcting Output Coding, that significantly improve classifiers like decision trees, rule learners, or neural networks. Unfortunately, t hese combining methods do not improve the nearest neighbor classifier. In this paper, we present MFS, a combining algorithm designed to improve the accuracy of the nearest neighbor (NN) classifier. MFS combines multiple NN classifiers each using only a random subset of features. The experimental results are encouraging: On 25 datasets from the UCI repository, MFS significantly outperformed several standard NN variants and was competitive with boosted decision trees. In additional experiments, we show that MFS is robust to irrelevant features, and is able to reduce both bias and variance components of error.

This study evaluated whether articulatory kinematics, inferred by articulatory phonology speech inversion neural networks, aligned with perceptual ratings of /ɹ/ and /s/ in the speech of children with speech sound disorders. Articulatory phonology vocal tract variables were inferred for 5,961 utterances from 118 children and three adults, aged 2.25-45 years. Perceptual ratings were standardized using the novel 5-point PERCEPT Rating Scale and training protocol. Two research questions examined if the articulatory patterns of inferred vocal tract variables aligned with the perceptual error category for the phones investigated (e.g., tongue tip is more anterior in dentalized /s/ productions than in correct /s/). A third research question examined if gradient PERCEPT Rating Scale scores predicted articulatory proximity to correct productions. Estimated marginal means from linear mixed models supported 17 of 18 /ɹ/ hypotheses, involving tongue tip and tongue body constrictions. For /s/, estimated marginal means from a second linear mixed model supported seven of 15 hypotheses, particularly those related to the tongue tip. A third linear mixed model revealed that PERCEPT Rating Scale scores significantly predicted articulatory proximity of errored phones to correct productions. Inferred vocal tract variables differentiated category and magnitude of articulatory errors for /ɹ/, and to a lesser extent for /s/, aligning with perceptual judgments. These findings support the clinical interpretability of speech inversion vocal tract variables and the PERCEPT Rating Scale in quantifying articulatory proximity to the target sound, particularly for /ɹ/. https://doi.org/10.23641/asha.31152061.

This paper describes an audio visual speech recognition (AVSR) system based on articulatory features (AF). It implements a tandem approach where artificial neural networks (ANN), in particular multi-layer perceptrons (MLP), are used as posterior probability estimators for transforming raw input data into the more abstract articulatory features. Such an approach is particularly well suited if relatively few training data are available, a situation which is typical for AVSR. In addition, the MLP feature extraction results and some analysis in terms of recognition accuracy and confusions are presented. Our AF-based AVSR system has been trained on the audio-visual speech corpus VIDTIMIT, which contains conversational speech based on a medium size vocabulary including more than 1200 words.

In this paper, we propose a novel framework to integrate articulatory features (AFs) into HMM- based ASR system. This is achieved by using posterior probabilities of different AFs (estimated by multilayer perceptrons) directly as observation features in Kullback-Leibler divergence based HMM (KL-HMM) system. On the TIMIT phoneme recognition task, the proposed framework yields a phoneme recognition accuracy of 72.4% which is comparable to KL-HMM system using posterior probabilities of phonemes as features (72.7%). Furthermore, a best performance of 73.5% phoneme recognition accuracy is achieved by jointly modeling AF probabilities and phoneme probabilities as features. This shows the efficacy and flexibility of the proposed approach.

A fully automated method for coronary calcification detection from non-contrast-enhanced, ECG-gated multi-slice computed tomography (CT) data is presented. Candidates for coronary calcifications are extracted by thresholding and component labeling. These candidates include coronary calcifications, calcifications in the aorta and in the heart, and other high-density structures such as noise and bone. A dedicated set of 64 features is calculated for each candidate object. They characterize the object's spatial position relative to the heart and the aorta, for which an automatic segmentation scheme was developed, its size and shape, and its appearance, which is described by a set of approximated Gaussian derivatives for which an efficient computational scheme is presented. Three classification strategies were designed. The first one tested direct classification without feature selection. The second approach also utilized direct classification, but with feature selection. Finally, the third scheme employed two-stage classification. In a computationally inexpensive first stage, the most easily recognizable false positives were discarded. The second stage discriminated between more difficult to separate coronary calcium and other candidates. Performance of linear, quadratic, nearest neighbor, and support vector machine classifiers was compared. The method was tested on 76 scans containing 275 calcifications in the coronary arteries and 335 calcifications in the heart and aorta. The best performance was obtained employing a two-stage classification system with a k-nearest neighbor (k-NN) classifier and a feature selection scheme. The method detected 73.8% of coronary calcifications at the expense of on average 0.1 false positives per scan. A calcium score was computed for each scan and subjects were assigned one of four risk categories based on this score. The method assigned the correct risk category to 93.4% of all scans.

The characteristics of society in choosing contraceptive methods are also the crucial factors for the government to prepare the family planning services needed at Bulakamba District, Brebes Regency, Central Java. In this case, a classification process needs to be done to assist the process of classifying the characteristics of society in the selection of contraceptive methods. Multinomial Logistic Regression classification is good in exploring data information meanwhile Fuzzy K Nearest Neighbor (FK-NN) classification is good for handling big data and noise. These two methods used in this study because they are relevant to the data applied and will be compared their classification accuracy through APER and Press's Q calculations.The classification accuracy results obtained based on the APER calculation for Multinomial Logistic Regression is 88,25% and Fuzzy K Nearest Neighbor (FK-NN) is 88,92%. Meanwhile, the Press's Q value of both methods are 9600,945 and 9518,014 greater than χ 2𝛼,1 which is 3,841, so that it is statistically accurate. Based on the results obtained, it can be concluded that Multinomial Logistic Regression classification method has a better classification accuracy than Fuzzy K Nearest Neighbor (FK-NN) method.

Forensic human identification with targeted microbiome markers using nearest neighbor classification

This paper is about a variant of k nearest neighbor classification on large many-class high dimensional datasets.K nearest neighbor remains a popular classification technique, especially in areas such as computer vision, drug activity prediction and astrophysics. Furthermore, many more modern classifiers, such as kernel-based Bayes classifiers or the prediction phase of SVMs, require computational regimes similar to k-NN. We believe that tractable k-NN algorithms therefore continue to be important.This paper relies on the insight that even with many classes, the task of finding the majority class among the k nearest neighbors of a query need not require us to explicitly find those k nearest neighbors. This insight was previously used in (Liu et al., 2003) in two algorithms called KNS2 and KNS3 which dealt with fast classification in the case of two classes. In this paper we show how a different approach, IOC (standing for the International Olympic Committee) can apply to the case of n classes where n > 2.IOC assumes a slightly different processing of the datapoints in the neighborhood of the query. This allows it to search a set of metric trees, one for each class. During the searches it is possible to quickly prune away classes that cannot possibly be the majority.We give experimental results on datasets of up to 5.8 x 105 records and 1.5 x 103 attributes, frequently showing an order of magnitude acceleration compared with each of (i) conventional linear scan, (ii) a well-known independent SR-tree implementation of conventional k-NN and (iii) a highly optimized conventional k-NN metric tree search.