CLOSE—A Data-Driven Approach to Speech Separation

This paper studies single-channel speech separation, assuming unknown, arbitrary temporal dynamics for the speech signals to be separated. A data-driven approach is described, which matches each mixed speech segment against a composite training segment to separate the underlying clean speech segments. To advance the separation accuracy, the new approach seeks and separates the longest mixed speech segments with matching composite training segments. Lengthening the mixed speech segments to match reduces the uncertainty of the constituent training segments, and hence the error of separation. For convenience, we call the new approach Composition of Longest Segments, or CLOSE. The CLOSE method includes a data-driven approach to model long-range temporal dynamics of speech signals, and a statistical approach to identify the longest mixed speech segments with matching composite training segments. Experiments are conducted on the Wall Street Journal database, for separating mixtures of two simultaneous large-vocabulary speech utterances spoken by two different speakers. The results are evaluated using various objective and subjective measures, including the challenge of large-vocabulary continuous speech recognition. It is shown that the new separation approach leads to significant improvement in all these measures.

Process Modelling for Control of Product Wall Thickness in Thermoforming

The present paper describes the results of an investigation into the modelling of plug assisted thermoforming. The objective of this work was to improve the finite element modelling of thermoforming through an enhanced understanding of the physical elements underlying the process. Experiments were carried out to measure the effects on output of changes in major parameters and simultaneously simple finite element models were constructed. The experimental results show that the process creates conflicting and interrelated contact friction and heat transfer effects that largely dictate the final wall thickness distribution. From the simulation work it was demonstrated that a high coefficient of friction and no heat transfer can give a good approximation of the actual wall thickness distribution. However, when conduction was added to the model the results for lower friction values were greatly improved. It was concluded that further work is necessary to provide realistic measurements and models for contact effects in thermoforming.

The role of plug design in determining wall thickness distribution in thermoforming

An experimental investigation has been carried out into the effects of changes in plug design on the wall thickness distribution of thermoformed products. Plugs were machined with a series of geometrical variations and their effects on the process were measured. The overall results show that the plug has a crucial role in controlling the wall thickness distribution in thermoforming. Larger plugs tend to distribute more material to the base of the product, but the introduction of a small sidewall taper, base radius, or a reduction in plug diameter tend to lead to more balanced distributions. However, larger changes in any of the variables tend to destroy these benefits. It has also been demonstrated that the frictional and thermal properties of the plug are important in determining the deformation response of the sheet material. There is a clear evidence of slip in the sheet during plug contact and, although the cooling effect of the plug appears to be minimal, cooling in the highly deformed regions away from the plug appears to be a significant factor.