Flow-based generative models are composed of invertible transformations between two random variables the same dimension. Therefore, flow-based cannot be adequately trained if dimension data distribution does not match that underlying target distribution. In this paper, we propose SoftFlow, a probabilistic framework for training normalizing flows on manifolds. To sidestep mismatch problem, SoftFlow estimates conditional perturbed input instead learning directly. We experimentally show can...

Recently, generative adversarial networks (GANs) have been successfully applied to speech enhancement. However, there still remain two issues that need be addressed: (1) GAN-based training is typically unstable due its non-convex property, and (2) most of the conventional methods do not fully take advantage characteristics, which could result in a sub-optimal solution. In order deal with these problems, we propose progressive generator can handle multi-resolution fashion. Additionally,...

In plant factories, light is fully controllable for crop production but involves a cost. For efficient lighting, use efficiency (LUE) should be considered as part of environment design. The objectives this study were to evaluate and interpret the interception, photosynthetic rate, LUE lettuces under electrical lights using ray-tracing simulation. architecture model was constructed by 3D scanning, simulation used interception photosynthesis. evaluation reliability, measured intensities rates...

Over the recent years, various self-supervised embedding learning methods for deep speaker verification were proposed. The performance of framework highly depends on data augmentation technique, but due to sensitive nature information within speech signal, most training relies simple augmentations such as additive noise or simulated reverberation. Thus while conventional systems can yield minimum within-utterance variability, their capability generalize out-of-set utterance is limited. In...

Ever since the deep neural network (DNN)-based acoustic model appeared, recognition performance of automatic speech has been greatly improved. Due to this achievement, various researches on DNN-based technique for noise robustness are also in progress. Among these approaches, noise-aware training (NAT) which aims improve inherent DNN using estimates shown remarkable performance. However, despite great performance, we cannot be certain whether NAT is an optimal method sufficiently utilizing...

인공광 식물공장은 조명에 많은 전력이 소요되므로 광이용효율의 향상은 필수적이다. 본 연구에서는 식물공장의 LED 광원 에 산란 유리를 사용함으로써 광이용효율을 향상하고자 하였다. 실험에는 Haze factor 40%와 80%의 유리가 적용된 두 가지 처리구와 사용하지 않은 대조구를 두었다. 3-D 광선추적기법을 이용하여 상추 작물의 군락 광분포를 분석하였으며 광합성률은 밀폐 아크릴 챔버를 측정하였다. 각 처리구 별로 16주의 상추를 수경재배 방식으로 28일간 재배하여 생장량을 비교하였다. 시뮬레이션 결과, 모든 처리구에서 상단부에 광량이 집중되었고 대조구에서 작물 상부에 광량 핫스팟이 발생하며 처리구에 비하여 광이용효율이 감소하였다. 총 광 흡수량은 가장 높았으나 유효 더 높았고 산란광의 비율이 높은 처리에서 높게 나타났다. 광합성률과 생장량은 사용한 대조구에 비해 결과적으로 유리의 이용을 통해 상추의 내 광분포가 개선되었고 광합성률 및 생장량이 증가하여 향상되는 것으로 나타났다 .

Although spectrum conversion films are used to improve the photosynthetic efficiency and, ultimately, crop growth, effects of modified on traits in plants have not yet been sufficiently reported. The objective this study was investigate changes performance and chlorophyll fluorescence sweet pepper (Capsicum annuum L.) under a green-to-red (GtR) film. GtR-modified increased dry mass decreased petiole length. light-response curves were significantly improved, gap maximum rates over time after...

In this paper, we propose a simple but powerful unsupervised learning method for speaker recognition, namely Contrastive Equilibrium Learning (CEL), which increases the uncertainty on nuisance factors latent in embeddings by employing uniformity loss. Also, to preserve discriminability, contrastive similarity loss function is used together. Experimental results showed that proposed CEL significantly outperforms state-of-the-art verification systems and best performing model achieved 8.01%...

This work focuses on the problem of detecting fake audio clips. To improve current spoofing detection models, we propose a selection multiple augmentations spe-cially designed to resemble attacks. These are experimentally found be very useful and using them achieves notable performance 2.8% EER ASVspoof 2019 challenge evaluation set. Unlike widely employed acoustic features, in this paper explore use Mel-spectrogram image features employ vari-ous codecs achieve robustness codec transmission...

Over the recent years, various deep learning-based embedding methods have been proposed and shown impressive performance in speaker verification. However, as most of classical techniques, are known to suffer from severe degradation when dealing with speech samples different conditions (e.g., recording devices, emotional states). In this paper, we propose a novel fully supervised training method for extracting vector disentangled variability caused by nuisance attributes. The framework was...

Extraction of a speaker embedding vector plays an important role in deep learning-based verification. In this contribution, to extract discriminant utterance level embeddings, we propose hybrid neural network that employs both cross- and self-module attention pooling mechanisms. More specifically, the proposed system incorporates 2D-Convolution Neural Network (CNN)-based feature extraction module cascade with frame-level network, which is composed fully Time Delay (TDNN) TDNN-Long Short Term...

Although plant responses to artificial lighting spectra often produce abnormal morphogenesis and reduced productivity, no quantification method determine how plants perceive respond light has been available. Our objective in this study was test whether a plant’s spectral perception can be quantified using the absorption of its major photoreceptors, phytochrome, cryptochrome, phototropin. We developed an solar lamp three different sources, based on high-pressure sodium lamp, fluorescent red...

In recent years, various flow-based generative models have been proposed to generate high-fidelity waveforms in real-time. However, these require either a well-trained teacher network or number of flow steps making them memory-inefficient. this paper, we propose novel model called WaveNODE which exploits continuous normalizing for speech synthesis. Unlike the conventional models, places no constraint on function used operation, thus allowing usage more flexible and complex functions....

In this paper, we propose a new hybrid system for extracting speaker embedding vector. More specifically, the proposed employs multi-level global-local statistics pooling method in order to aggregate information within short time-span and utterance-level context. evaluate system, set of experiments on NIST SRE 2016, Short-duration verification (SdSV) Challenge 2021, VoxCeleb datasets were conducted, network was able outperform conventional approaches trained same dataset. Moreover, our...