Search

Abstract Non-parallel data voice conversion (VC) has achieved considerable breakthroughs due to self-supervised pre-trained representation (SSPR) being used in recent years. Features extracted by the pre-trained model are expected to contain more content information. However, in common VC with SSPR, there is no special implementation to remove speaker information in the content representation extraction by SSPR, which prevents further purification of the speaker information from SSPR representation. Moreover, in conventional VC, Mel-spectrogram is often selected as the reconstructed acoustic feature, which is not consistent with the input of the content encoder and results in some information lost. Motivated by the above, we proposed W2VC to settle the issues. W2VC consists of three parts: (1) We reconstruct feature from WavLM representation (WLMR) that is more consistent with the input of content encoder; (2) Connectionist temporal classification (CTC) is used to align content representation and text context from phoneme level, content encoder plus gradient reversal layer (GRL) based speaker classifier are used to remove speaker information in the content representation extraction; (3) WLMR-based HiFi-GAN is trained to convert WLMR to waveform speech. VC experimental results show that GRL can purify well the content information of the self-supervised model. The GRL purification and CTC supervision on the content encoder are complementary in improving the VC performance. Moreover, the synthesized speech using the WLMR retrained vocoder achieves better results in both subjective and objective evaluation. The proposed method is evaluated on the VCTK and CMU databases. It is shown the method achieves 8.901 in objective MCD, 4.45 in speech naturalness, and 3.62 in speaker similarity of subjective MOS score, which is superior to the baseline.

Abstract This paper tackles the front‐back disambiguity problem in speaker localization when the audio signals are captured by a symmetric microphone array. To this end, a deep neural network is proposed with an attention‐based mechanism designed to assign different weights to features obtained from individual microphones. For support, a real dataset with synchronized multichannel audio signals captured by a large linear microphone array is introduced, along with manual annotations. The experimental results demonstrate the effectiveness of the proposed method over the other approaches. In particular, more than 50% reduction in Equal Error Rate (EER) is achieved when comparing with the single‐channel case. The designed multi‐channel self‐attention mechanism also brings further improvements. The dataset and source code will be released.

Energy system information for electricity access planning such as the locations and connectivity of electricity transmission and distribution towers—termed the power grid—is often incomplete, outdated, or altogether unavailable. Furthermore, conventional means for collecting this information is costly and limited. We propose to automatically map the grid in overhead remotely sensed imagery using an deep learning approach. Toward this goal, we develop and publicly release a large dataset (263 km$^2$) of overhead imagery with ground-truth for the power grid—to our knowledge, this is the first dataset of its kind in the public domain. Additionally, we propose scoring metrics and baseline algorithms for two grid-mapping tasks: 1) tower recognition and 2) power line interconnection (i.e., estimating a graph representation of the grid). We hope the availability of the training data, scoring metrics, and baselines will facilitate rapid progress on this important problem to help decision-makers address the energy needs of societies around the world.

Surgery is currently the most appropriate treatment for early-stage non-small cell lung cancer (NSCLC). Increasing unilateral or bilateral multiple primary lung cancer being found, segmentectomy has attracted wide attention for its unique advantages in the treatment for such tumors. Ground glass opacity dominant early-stage NSCLC is associated with a good prognosis and can be cured by segmentectomy, however, the treatment of solid-dominant NSCLC remains controversial owing to the invasive nature. With the in-depth study on the lymph node metastasis pathway, radiological characteristics and molecular biology of NSCLC, a large part of solid nodules with certain characteristics can also be cured by segmentectomy. This paper reviews the research status and progress about the indication of segmentectomy.

Abstract Objective It has been confirmed that the micropapillary (MP) pattern is a poor prognostic factor after resection of lung adenocarcinoma (ADC), but the proportion of the MP component as a prognostic criterion is still controversial. Hence, a meta-analysis was performed to evaluate whether the presence of an MP component has equal prognostic power as the MP predominant subtype. Methods Literature retrieval was performed in the MEDLINE, EMBASE, and Cochrane databases until December 23, 2019. Eligible studies were selected based on the inclusion and exclusion criteria. The included studies were divided into two subgroups, the MP component subgroup and the MP predominant subgroup, according to the proportion of the MP pattern to analyse the effect of this pattern on disease-free survival (DFS) and overall survival (OS). The hazard ratio (HR) and 95% confidence interval (CI) were extracted from each study. Review Manager 5.3 was used for statistical analyses. Results Finally, 10 studies, including a total of 4934 lung ADC patients, were included in this meta-analysis. Our results indicated a significantly worse pooled DFS (HR 1.62, 95% CI 1.20–2.21) and OS (HR 1.53, 95% CI 1.19–1.96) in the subgroup of MP predominant subtype patients. The pooled DFS (HR 1.80, 95% CI 1.45–2.85) and OS (HR 2.26, 95% CI 1.46–3.52) were also poor in the subgroup of patients with the presence of an MP component. Conclusions Both the presence of an MP component and the MP predominant subtype are related to poor DFS and OS after lung ADC resection and represent adverse prognostic factor for lung ADC patients. However, there are some limitations in this meta-analysis, and quantitative stratification based on the proportion of the MP component is needed to explore its effect on prognosis of lung ADC patients in the future.

Abstract In order to improve the performance of hand-crafted features to detect playback speech, two discriminative features, constant-Q variance-based octave coefficients and constant-Q mean-based octave coefficients, are proposed for playback speech detection in this work. They rely on our findings that variance-based modified log magnitude spectrum and mean-based modified log magnitude spectrum can enhance the discriminative power between genuine speech and playback speech. Then constant-Q variance-based octave coefficients (constant-Q mean-based octave coefficients) can be obtained by combining variance-based modified log magnitude spectrum (mean-based modified log magnitude spectrum), octave segmentation, and discrete cosine transform. Finally, constant-Q variance-based octave coefficients and constant-Q mean-based octave coefficients are evaluated on ASVspoof 2017 corpus version 2.0 and ASVspoof 2019 physical access, respectively. Experimental results show that variance-based modified log magnitude spectrum and mean-based modified log magnitude spectrum can produce discriminative features toward playback speech. Further results on the two databases show that constant-Q variance-based octave coefficients and constant-Q mean-based octave coefficients can perform better than some common features, such as mel frequency cepstral coefficients and constant-Q cepstral coefficients.

Conventional speaker verification systems become frail or incompetent while facing attack from spoofed speech. Presently many anti-spoofing countermeasures have been studied for automatic speaker verification. It has been known that the salient feature is of a more important role rather than the selection of classifiers in the current research field of spoofing detection. The effectiveness of constant-Q transform (CQT) has been demonstrated for anti-spoofing feature analysis in many research literatures on automatic speaker verification. On the basis of CQT-based information sub-features, i.e. octave-band principal information (OPI), full-band principal information (FPI), short-term spectral statistics information (STSSI) and magnitude-phase energy information (MPEI), three concatenated features are proposed by investigating their information complementarity in this paper, the first one is constant-Q statistics-plus-principal information coefficients (CQSPIC) by combining OPI, FPI and STSSI; the second one is constant-Q energy-plus-principal information coefficients (CQEPIC) by combining OPI, FPI and MPEI and the third one is constant-Q energy-statistics-principal information coefficients (CESPIC) by combining OPI, FPI, MPEI and STSSI. In this paper, we set up deep neural network (DNN) classifiers for evaluation of the proposed features. Experiments show that the proposed features can outperform some commonly used features meanwhile the proposed systems give better or comparable performance comparing with state-of-the-art performance on ASVspoof 2019 logical access and physical access corpus.

Abstract Objective Whether segmentectomy can be used to treat radiologically determined pure solid or solid-dominant lung cancer remains controversial owing to the invasive pathologic characteristics of these tumors despite their small size. This meta-analysis compared the oncologic outcomes after lobectomy and segmentectomy regarding relapse-free survival (RFS) and overall survival (OS) in patients with radiologically determined pure solid or solid-dominant clinical stage IA non-small cell lung cancer (NSCLC). Methods A literature search was performed in the MEDLINE, EMBASE, and Cochrane Central databases for information from the date of database inception to March 2019. Studies were selected according to predefined eligibility criteria. The hazard ratio (HR) and associated 95% confidence interval (CI) were extracted or calculated as the outcome measure for data combining. Results Seven eligible studies published between 2014 and 2018 enrolling 1428 patients were included in the current meta-analysis. Compared with lobectomy, segmentectomy had a significant benefit on the RFS of radiologically determined pure solid or solid-dominant clinical stage IA NSCLC patients (combined HR: 1.46; 95% CI, 1.05–2.03; P = 0.024) and there were no significant differences on the OS of these patients (HR: 1.52; 95% CI, 0.95–2.43; P = 0.08). Conclusions Segmentectomy leads to lower survival than lobectomy for clinical stage IA NSCLC patients with radiologically determined pure solid or solid-dominant tumors. Moreover, applying lobectomy to clinical stage IA NSCLC patients with radiologically determined pure solid or solid-dominant tumors (≤2 cm) could lead to an even bigger survival advantage. However, there are some limitations in the present study, and more evidence is needed to support the conclusion.

The pathophysiology of cardiac hypertrophy is complex and multifactorial. Both the store-operated Ca2+ entry (SOCE) and excessive autophagy are the major causative factors for pathological cardiac hypertrophy. However, it is unclear whether these two causative factors are interdependent. In this study, we examined the functional role of SOCE and Orai1 in angiotensin II (Ang II)-induced autophagy and hypertrophy using in vitro neonatal rat cardiomyocytes (NRCMs) and in vivo mouse model, respectively. We show that YM-58483 or SKF-96365 mediated pharmacological inhibition of SOCE, or silencing of Orai1 with Orail-siRNA inhibited Ang II-induced cardiomyocyte autophagy both in vitro and in vivo. Also, the knockdown of Orai1 attenuated Ang II-induced pathological cardiac hypertrophy. Together, these data suggest that Ang II promotes excessive cardiomyocyte autophagy through SOCE/Orai1 which can be the prime contributing factors in cardiac hypertrophy.

In this article, we study oscillatory properties of solutions to neutral nonlinear impulsive hyperbolic partial differential equations with several delays. We establish sufficient conditions for oscillation of all solutions.

Previously we have shown that kallikreins (klks) play a renoprotective role in nephrotoxic serum induced nephritis. In this study, we have used mesenchymal stem cells (MSCs) as vehicles to deliver klks into the injured kidneys and have measured their therapeutic effect on experimental antibody induced nephritis and lupus nephritis. Human KLK-1 (hKLK1) gene was transduced into murine MSCs using a retroviral vector to generate a stable cell line, hKLK1-MSC, expressing high levels of hKLK1. 129/svj mice subjected to anti-GBM induced nephritis were transplanted with 10(6) hKLK1-MSCs and hKLK1 expression was confirmed in the kidneys. Compared with vector-MSCs injected mice, the hKLK1-MSCs treated mice showed significantly reduced proteinuria, blood urea nitrogen (BUN) and ameliorated renal pathology. Using the same strategy, we treated lupus-prone B6.Sle1.Sle3 bicongenic mice with hKLK1-MSCs and demonstrated that hKLK1-MSCs delivery also attenuated lupus nephritis. Mechanistically, hKLK1-MSCs reduced macrophage and T-lymphocyte infiltration into the kidney by suppressing the expression of inflammation cytokines. Moreover, hKLK1 transduced MSCs were more resistant to oxidative stress-induced apoptosis. These findings advance genetically modified MSCs as potential gene delivery tools for targeting therapeutic agents to the kidneys in order to modulate inflammation and oxidative stress in lupus nephritis.