Improving Throat Microphone Speech Recognition by Joint Analysis of Throat and Acoustic Microphone Recordings

We present a new framework for joint analysis of throat and acoustic microphone (TAM) recordings to improve throat microphone only speech recognition. The proposed analysis framework aims to learn joint sub-phone patterns of throat and acoustic microphone recordings through a parallel branch HMM structure. The joint sub-phone patterns define temporally correlated neighborhoods, in which a linear prediction filter estimates a spectrally rich acoustic feature vector from throat feature vectors. Multimodal speech recognition with throat and throat-driven acoustic features significantly improves throat-only speech recognition performance. Experimental evaluations on a parallel TAM database yield benchmark phoneme recognition rates for throat-only and multimodal TAM speech recognition systems as 46.81% and 60.69%, respectively. The proposed throat-driven multimodal speech recognition system improves phoneme recognition rate to 52.58%, a significant relative improvement with respect to the throat-only speech recognition benchmark system.     

Factors influencing the crushing strength of some Aegean sands

Engineering properties of sands mainly depend on the integrity of the particles, which in turn has a strong bearing on their crushing strength. Seven different Aegean sands were tested for mineralogy, particle shape, size and specific gravity and the influence of aspect ratio, particle composition, particle shape and size on the crushing strength was examined. As the Aegean sands have a small range of sphericity and roundness, crushing strength tests were also performed on five Anatolian sands. A multiple regression analysis was carried out and an equation proposed to determine the crushing strength value of the Aegean sands. The computed values were found to be in good agreement with those obtained from the experimental investigations. It is concluded that the equation is sufficiently accurate to be a useful, time- and cost-effective way of obtaining crushing strength estimations at the preliminary stage of site investigations.     

Discriminative analysis of lip motion features for speaker identification and speech-reading.

There have been several studies that jointly use audio, lip intensity, and lip geometry information for speaker identification and speech-reading applications. This paper proposes using explicit lip motion information, instead of or in addition to lip intensity and/or geometry information, for speaker identification and speech-reading within a unified feature selection and discrimination analysis framework, and addresses two important issues: 1) Is using explicit lip motion information useful, and, 2) if so, what are the best lip motion features for these two applications? The best lip motion features for speaker identification are considered to be those that result in the highest discrimination of individual speakers in a population, whereas for speech-reading, the best features are those providing the highest phoneme/word/phrase recognition rate. Several lip motion feature candidates have been considered including dense motion features within a bounding box about the lip, lip contour motion features, and combination of these with lip shape features. Furthermore, a novel two-stage, spatial, and temporal discrimination analysis is introduced to select the best lip motion features for speaker identification and speech-reading applications. Experimental results using an hidden-Markov-model-based recognition system indicate that using explicit lip motion information provides additional performance gains in both applications, and lip motion features prove more valuable in the case of speech-reading application.     

Multimodal speaker identification using an adaptive classifier cascade based on modality reliability

We present a multimodal open-set speaker identification system that integrates information coming from audio, face and lip motion modalities. For fusion of multiple modalities, we propose a new adaptive cascade rule that favors reliable modality combinations through a cascade of classifiers. The order of the classifiers in the cascade is adaptively determined based on the reliability of each modality combination. A novel reliability measure, that genuinely fits to the open-set speaker identification problem, is also proposed to assess accept or reject decisions of a classifier. A formal framework is developed based on probability of correct decision for analytical comparison of the proposed adaptive rule with other classifier combination rules. The proposed adaptive rule is more robust in the presence of unreliable modalities, and outperforms the hard-level max rule and soft-level weighted summation rule, provided that the employed reliability measure is effective in assessment of classifier decisions. Experimental results that support this assertion are provided.     

Dynamic Aspects of Chip Formation

A multimass chain model of chip formation in the secondary deformation zone is proposed. The influence of the frictional conditions and cutting conditions on the group motion of the chip elements over the cutter’s front surface is shown. The dynamics of the motion of the chip’s free end has a significant influence on its fracture. In particular, the mass ratio of the last elements involved in the group motion (and positioned on the contact line at the cutter’s front surface) and the free end of the chip gives rise to considerable dynamic splitting forces between the elements. The proposed approach offers broad scope for analysis of the conditions of chip formation, including conditions with intensification of the cutting process by electrical pulses.     

Artificial neural network models for predicting soil thermal resistivity

Thermal properties of soils are of great importance in view of the modern trends of utilizing the subsurface for transmission of either heated fluids or high power currents. For these situations, it is essential to estimate the resistance offered by the soil mass in dissipating the heat generated through it. Several investigators have tried to develop mathematical and theoretical models to estimate soil thermal resistivity. However, it is evident that these models are not efficient enough to predict accurate thermal resistivity of soils. This is mainly due to the fact that thermal resistivity of soils is a complex phenomenon that depends upon various parameters viz., type of the soil, particle size distribution and its compaction characteristics (i.e., dry density and moisture content). To overcome this, Artificial Neural Network (ANN) models, which are based on experimentally obtained thermal resistivity values for clay, silt, silty-sand, fine- and coarse-sands, have been developed. Incidentally, these soils are the most commonly encountered soils in nature and exhibit entirely different characteristics. The thermal resistivity of these soils, corresponding to their different compaction states, was obtained with the help of a laboratory thermal probe and compared vis-à-vis those obtained from the ANN model. The thermal resistivity of these soils obtained from ANN models and experimental investigations are found to match extremely well. The performance indices such as coefficient of determination, root mean square error, mean absolute error, and variance account for were used to control the performance of the prediction capacity of the models developed in this study. In addition to this, thermal resistivity of these soils obtained from ANN models were compared with those computed from the empirical relationships reported in the literature and were found to be superior. The study demonstrates the utility and efficiency of the ANN model for estimating thermal resistivity of soils.     

Skew Minimization Problem with Possible Sink Displacement

In applications such as design of integrated circuits (chips), it is sometimes required to connect the terminals (receivers) and the central vertex (source) by a weight-minimal tree where the signal delays between the source and the terminals are the same or differ by a minimal value. Simple necessary conditions for existence of the desired tree, heuristic rules for displacement of the terminals for which these conditions are not satisfied, and a new polynomial algorithm to determine an approximate solution were presented.