In this paper, we propose a novel speech dereverberation framework that utilizes deep neural network (DNN)-based spectrum estimation to construct linear inverse filters. The proposed dereverberation framework is based on the state-of-the-art inverse filter estimation algorithm called weighted prediction error (WPE) algorithm, which is known to effectively reduce reverberation and greatly boost the ASR performance in various conditions. In WPE, the accuracy of the inverse filter estimation, and thus the deverberation performance, is largely dependent on the estimation of the power spectral density (PSD) of the target signal. Therefore, the conventional WPE iteratively performs the inverse filter estimation, actual dereverberation and the PSD estimation to gradually improve the PSD estimate. However, while such iterative procedure works well when sufficiently long acoustically-stationary observed signals are available, WPE's performance degrades when the duration of observed/accessible data is short, which typically is the case for real-time applications using online block-batch processing with small batches. To solve this problem, we incorporate the DNN-based spectrum estimator into the framework of WPE, because a DNN can estimate the PSD robustly even from very short observed data. We experimentally show that the proposed framework outperforms the conventional WPE, and improves the ASR performance in real noisy reverberant environments in both single-channel and multichannel cases.

Abstract. Previous studies have suggested that ozonated water is safe and possesses antibacterial effects for treatment of experimental peritonitis rats. In this study, we evaluated the anti-inflammatory effects of ozonated water that was intraperitoneally injected into an experimental inflammatory mouse model. The concentrations of dissolved ozone decreased constantly and lineally, while the half-life of dissolved ozone was 36.8±2.7 min (27˚C). The 10-ppm ozonated water was injected intraperitoneally into mice with lipopolysaccharide (LPS)-induced acute inflammation. The results showed that the intraperitoneal injection of ozonated water decreased the levels of tumor necrosis factor-α (TNF-α) and increased the activity of superoxide dismutase (SOD). The results suggest that ozonated water has anti-inflammatory properties and is a potential therapeutic option for acute inflammation.

Ozonated water is easier to handle than ozone gas. However, there have been no previous reports on the biological effects of ozonated water. We conducted a study on the safety of ozonated water and its anti-tumor effects using a tumor-bearing mouse model and normal controls. Local administration of ozonated water (208 mM) was not associated with any detrimental effects in normal tissues. On the other hand, local administration of ozonated water (20.8, 41.6, 104, or 208 mM) directly into the tumor tissue induced necrosis and inhibited proliferation of tumor cells. There was no significant difference in the number of terminal deoxynucleotidyl transferase-mediated deoxyuridine triphosphate-biotin nick-end labeling (TUNEL)-positive cells following administration of ozonated water. The size of the necrotic areas was dependent on the concentration of ozonated water. These results indicate that ozonated water does not affect normal tissue and damages only the tumor tissue by selectively inducing necrosis. There is a possibility that it exerts through the production of reaction oxygen species (ROS). In addition, the induction of necrosis rather than apoptosis is very useful in tumor immunity. Based on these results, we believe that administration of ozonated water is a safe and potentially simple adjunct or alternative to existing antineoplastic treatments.

End-to-end deep learning has become a popular framework for automatic speech recognition (ASR) tasks, and it has proven itself to be a powerful solution. Unfortunately, network structures commonly have millions of parameters, and large computational resources are required to make this approach feasible for training and running such networks. Moreover, many applications still prefer lightweight models of ASR that can run efficiently on mobile or wearable devices. To address this challenge, we propose an approach that can reduce the number of ASR parameters. Specifically, we perform Tensor-Train decomposition on the weight matrix of the gated recurrent unit (TT-GRU) in the end-to-end ASR framework. Experimental results on LibriSpeech data reveal that the compressed ASR with TT-GRU can maintain good performance while greatly reducing the number of parameters.

No abstract