Recently, Flexible pressure sensor-based Throat Microphones (FTM) have attracted more attention in noise-robust speaker recognition and are promising for helping people with specific dysarthria to complete speaker recognition. FTM has outstanding flexibility compared with Hard Throat Microphones (HTM) and noise-robustness compared with Close-talk microphones (CM). However, speaker recognition for FTM is still an open task awaiting exploration since FTM has degradation problems and a lack of data sets. To tackle these two obstacles, referring to feature mapping methods for HTM, we introduce an FTM-oriented supervised contrastive learning (FTMSCL) method. An FTM speech data set is collected, then a contrastive loss function is designed to avoid the feature mapping methods' problems and effectively leverage label information from this data set. Furthermore, a critical parameter margin in this loss and several data augmentations for FTM are investigated. Experimental results show that, with no need for CM data, FTMSCL can achieve a False Acceptance Rate (FAR) of 2.97% and a False Rejection Rate (FRR) of 2.83%, which outperforms a conventional End-to-End one and an advanced feature mapping one significantly. Moreover, the best FAR and FRR of our FTMSCL method are only 0.86% and 0.83% higher than the best one using clean CM data.

BACKGROUND: Comutation plot is a widely used visualization method to deliver a global view of the mutation landscape of large-scale genomic studies. Current tools for creating comutation plot are either offline packages that require coding or online web servers with varied features. When a package is used, it often requires repetitive runs of code to adjust a single feature that might only be a few clicks in a web app. But web apps mostly have limited capacity for customization and cannot handle very large genomic files. RESULTS: To improve on existing tools, we identified features that are most frequently adjusted in creating a plot and incorporate them in Comut-viz that interactively filters and visualizes mutation data as downloadable plots. It includes colored labels for numeric metadata, a preloaded palette for changing colors and two input boxes for adjusting width and height. It accepts standard mutation annotation format (MAF) files as input and can handle large MAF files with more than 200 k rows. As a front-end only app, Comut-viz guarantees privacy of user data and no latency in the analysis. CONCLUSIONS: Comut-viz is a highly responsive and extensible web app to make comutation plots. It provides customization for frequently adjusted features and accepts large genomic files as input. It is suitable for genomic studies with more than a thousand samples.

SUMMARY: We present HoPhage (Host of Phage) to identify the host of a given phage fragment from metavirome data at the genus level. HoPhage integrates two modules using a deep learning algorithm and a Markov chain model, respectively. HoPhage achieves 47.90% and 82.47% mean accuracy at the genus and phylum levels for ∼1-kb long artificial phage fragments when predicting host among 50 genera, representing 7.54-20.22% and 13.55-24.31% improvement, respectively. By testing on three real virome samples, HoPhage yields 81.11% mean accuracy at the genus level within a much broader candidate host range. AVAILABILITY AND IMPLEMENTATION: HoPhage is available at http://cqb.pku.edu.cn/ZhuLab/HoPhage/data/. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Noncoherent signal model is widely applied to angular superresolution in forward-looking imaging radar. However, relative phase influences the resolvability of the model in coherent applications. We propose a coherent signal model by replacing the antenna power pattern with the antenna radiation pattern. Experiment results have demonstrated significant resolution improvements for forward-looking imaging in coherent scanning radars.

The research of noisy speech emotion recognition is significant in practical applications. In this paper, samples reconstructed by Compressed Sensing are used in noisy speech emotion recognition and a secondary feature selection fusing filter feature selection algorithms is proposed to achieve more effective feature subset. Three reconstruction methods are adopted on the measurements of noisy samples to demonstrate the feasibility of sample reconstruction in speech emotion recognition. The negative impact of noise and the effectiveness of secondary feature selection are verified on five emotions in Berlin Database of Emotional Speech. Experimental results show that the combination of sample reconstruction and secondary feature selection can improve the emotion recognition accuracies of noisy samples to be close to or even higher than the accuracies of clean samples.