The performance of automatic speech recognition systems(ASR) degrades in the presence of noisy speech. This paper demonstrates that using electroencephalography (EEG) can help automatic speech recognition systems overcome performance loss in the presence of noise. The paper also shows that distillation training of automatic speech recognition systems using EEG features will increase their performance. Finally, we demonstrate the ability to recognize words from EEG with no speech signal on a limited English vocabulary with high accuracy.

Many studies have shown that articulatory features can significantly improve the performance of automatic speech recognition systems. Unfortunately, such features are not available at recognition time. There are two main approaches to solve this problem: a feature-based approach, the most popular example of which is the acoustic-to-articulatory inversion, where the missing articulatory features are generated from the speech signal, and a model-based approach, where articulatory information is embedded in the model structure and parameters in a way that allows recognition using only acoustic features. In this paper, we propose two new methods to integrate articulatory information into a phoneme recognition system. One of them is feature based, and the other is model based. In both cases, the underlying acoustic model (AM) is a deep neural networks-hidden Markov model (DNN-HMM) hybrid. In the feature-based method, the articulatory inversion DNN and the acoustic model DNN are trained jointly using a linear combination of their loss functions. In the model-based method, we utilize the generalized distillation framework to train the AM DNN. In this case, first, a teacher DNN is trained on both the acoustic and articulatory features, and then its outputs are used as additional targets during the AM DNN training with acoustic features only. A 7-fold cross-validation experiments using 42 speakers from the XRMB database showed that both the proposed methods provide about 22% to 25% performance improvement with respect to the DNN acoustic model trained with acoustic features only.

This study proposes a compact improved Koch snowflake fractal multi-broadband planar antenna for the second generation (2G) to the fifth generation (5G), WLAN and navigation wireless applications. The device is based on the principles and structural features of conventional Koch fractal monopole antenna elements, combined with the advantages of CPW-fed structure. The proposed antenna adapted two iterations Koch snowflake fractal structure radiator with one iteration Koch snowflake fractal slot inside and a novel hexagonal resonator on its back, which covers more than ten mobile applications in three broad frequency bands with a bandwidth of 76.4% (1.4-3.13 GHz) for DCS1800, TD-SCDMA, WCDMA, LTE33-41, Bluetooth, GPS, BDS, GLONSS, GALILEO, WLAN(802.11b/g/n) frequency bands, 12% (3.37-3.8 GHz) for LTE42,43 and WiMAX frequency bands, 36.3% (4.12-5.95 GHz) for WLAN(802.11a/n) and 5G systems frequency bands. The proposed antenna was fabricated on a 1.6 mm-thick G10/FR4 substrate with a dielectric constant of 4.4 and a size of 72*42mm2. The measurement results reveal that the omni-directional radiation patterns achieve a gain of 0.86-2.72 dBi. The good agreement between the measurement results and simulation validates the proposed design approach and satisfies the requirements for various wireless applications.

Smallcells have great potential to enhance cellular networks, complementing macrocells. Severe interference may occur, as smallcells are expected to be deployed and operated uncoordinatedly. However, existing resource management methods require significant overhead to suppress interference. We propose a new resource management approach which is able to mitigate the cross-tier and co-tier interference with substantially reduced overhead. The key idea is to categorize the smallcells into two regions based on a judiciously designed cross-tier interference criterion. Smallcells in the high-interference zone occupy orthogonal radio resources with the macrocell; smallcells in the other zone can reuse the resources that the macrocell is using. Another crucial aspect is that we formulate the resource sharing between the macrocell and smallcells in the low-interference zone to a multi-agent Q-learning process which assigns adequate transmit power levels in a decentralized manner to suppress the co/cross-tier interference. As a result, our approach is able to reduce the outage probabilities of macrocell users significantly to 0%, respectively, in a dense smallcell deployment (200 smallcells), as evidenced by simulation results.