Employing low-pass filtered temporal speech features for the training of ideal ratio mask in speech enhancement

The masking-based speech enhancement method pursues a multiplicative mask that applies to the spectrogram of input noise-corrupted utterance, and a deep neural network (DNN) is often used to learn the mask. In particular, the features commonly used for automatic speech recognition can serve as the input of the DNN to learn the well-behaved mask that significantly reduce the noise distortion of processed utterances. This study proposes to preprocess the input speech features for the ideal ratio mask (IRM)-based DNN by lowpass filtering in order to alleviate the noise components. In particular, we employ the discrete wavelet transform (DWT) to decompose the temporal speech feature sequence and scale down the detail coefficients, which correspond to the high-pass portion of the sequence. Preliminary experiments conducted on a subset of TIMIT corpus reveal that the proposed method can make the resulting IRM achieve higher speech quality and intelligibility for the babble noise-corrupted signals compared with the original IRM, indicating that the lowpass filtered temporal feature sequence can learn a superior IRM network for speech enhancement.

Mask Attention Networks: Rethinking and Strengthen Transformer

Transformer is an attention-based neural network, which consists of two sublayers, namely, Self-Attention Network (SAN) and Feed-Forward Network (FFN). Existing research explores to enhance the two sublayers separately to improve the capability of Transformer for text representation. In this paper, we present a novel understanding of SAN and FFN as Mask Attention Networks (MANs) and show that they are two special cases of MANs with static mask matrices. However, their static mask matrices limit the capability for localness modeling in text representation learning. We therefore introduce a new layer named dynamic mask attention network (DMAN) with a learnable mask matrix which is able to model localness adaptively. To incorporate advantages of DMAN, SAN, and FFN, we propose a sequential layered structure to combine the three types of layers. Extensive experiments on various tasks, including neural machine translation and text summarization demonstrate that our model outperforms the original Transformer.

Convolutional Neural Networks for Photo Features Extraction

يهدف البحث إلى تقديم دراسة مرجعيّة مفصلة عن استخدام الشبكات العصبونية الإلتفافية (CNNs) في استخراج الميزات (Features) من الصور. وسيتطرق البحث إلى التعريف بمعنى الميزات (Features) الخاصة بالصور وأهميتها في تطبيقات معالجة الصورة. وسيتم أيضاً التعريف بالشبكات العصبونية الإلتفافية (CNNs) وبنيتها و طريقة عملها وأنواع المقاربات والمنهجيات المستخدمة في تدريبها لاستخراج الميزات (Features) من الصور.

Viola's Disguise in Twelfth Night

This research is concerned with the issue of Viola's – the heroine - disguise in Twelfth Night. This young woman, who has closed the sad chapters of her life, decides to begin a new life, as if she had not known desperation by disguising as a young man named ''Cesario''. In this play, we notice that disguise turns from a simple device to a complicated one. At the beginning, Viola's aim is to protect herself in a country where she knows no one. But later, disguise comes to mean more than this. Cesario begins to dominate Viola, the characters and the play. He teaches everyone lessons in life about true love, faithfulness, sacrifice for the happiness of the others, and selflessness. Also, he removes the false masks of the others. Sebastian's appearance gives Viola the strength to admit, before all people, her real identity. This is what helps to solve a lot of ambiguity and questions for us and the characters. Viola's strength does not lie in her disguise as a man but in the way that she uses disguise and directs it.