Deep Convolutional Neural Networks (CNNs) are more powerful than Deep Neural Networks (DNN), as they are able to better reduce spectral variation in the input signal. This has also been confirmed experimentally, with CNNs showing improvements in word error rate (WER) between 4-12% relative compared to DNNs across a variety of LVCSR tasks. In this paper, we describe different methods to further improve CNN performance. First, we conduct a deep analysis comparing limited weight sharing and full weight sharing with state-of-the-art features. Second, we apply various pooling strategies that have shown improvements in computer vision to an LVCSR speech task. Third, we introduce a method to effectively incorporate speaker adaptation, namely fMLLR, into log-mel features. Fourth, we introduce an effective strategy to use dropout during Hessian-free sequence training. We find that with these improvements, particularly with fMLLR and dropout, we are able to achieve an additional 2-3% relative improvement in WER on a 50-hour Broadcast News task over our previous best CNN baseline. On a larger 400-hour BN task, we find an additional 4-5% relative improvement over our previous best CNN baseline.

No abstract

Mechanical stability of solder joints has influence on total reliability of electronic products. Mechanical stress affect on properties of solder joints during function, transport and storage of the electronic products. A method which is commonly used in electronic industry to classify the mechanical properties and reliability of solder joints is shear test. This article deals with the comparison of mechanical stability of soldered joints from the point of shear strength of soldered SMD resistors. Three types of solder pastes, one lead Sn62Pb36Ag2 and two lead free Sn42Bi58 and Sn96.5Ag3Cu0.5, were used for solder joint preparation. In addition, influence of solder paste amount on shear strength of soldered SMD component, were compared. The type of solder paste and its applied amount together with the temperature profile are interesting parameters from point of economic aspects, especially in case of mass production, where it is possible to achieve considerable savings. The results shows, that the amount of solder paste has influence on the share strength of solder joins. The best results, highest share strength, have been achieved for Sn96.5Ag3Cu0.5 solder paste.

No abstract

No abstract