Search

The objective of this paper is to combine multiple frame-level features into a single utterance-level representation considering pairwise relationship. For this purpose, we propose a novel graph attentive feature aggregation module by interpreting each frame-level feature as a node of a graph. The inter-relationship between all possible pairs of features, typically exploited indirectly, can be directly modeled using a graph. The module comprises a graph attention layer and a graph pooling layer followed by a readout operation. The graph attention layer first models the non-Euclidean data manifold between different nodes. Then, the graph pooling layer discards less informative nodes considering the significance of the nodes. Finally, the readout operation combines the remaining nodes into a single representation. We employ two recent systems, SE-ResNet and RawNet2, with different input features and architectures and demonstrate that the proposed feature aggregation module consistently shows a relative improvement over 10%, compared to the baseline., 5 pages, 1 figure, 6 tables, submitted to ICASSP 2022

Artefacts that differentiate spoofed from bona-fide utterances can reside in spectral or temporal domains. Their reliable detection usually depends upon computationally demanding ensemble systems where each subsystem is tuned to some specific artefacts. We seek to develop an efficient, single system that can detect a broad range of different spoofing attacks without score-level ensembles. We propose a novel heterogeneous stacking graph attention layer which models artefacts spanning heterogeneous temporal and spectral domains with a heterogeneous attention mechanism and a stack node. With a new max graph operation that involves a competitive mechanism and an extended readout scheme, our approach, named AASIST, outperforms the current state-of-the-art by 20% relative. Even a lightweight variant, AASIST-L, with only 85K parameters, outperforms all competing systems., 5 pages, 1 figure, 3 tables, submitted to ICASSP2022

Despite achieving satisfactory performance in speaker verification using deep neural networks, variable-duration utterances remain a challenge that threatens the robustness of systems. To deal with this issue, we propose a speaker verification system called RawNeXt that can handle input raw waveforms of arbitrary length by employing the following two components: (1) A deep layer aggregation strategy enhances speaker information by iteratively and hierarchically aggregating features of various time scales and spectral channels output from blocks. (2) An extended dynamic scaling policy flexibly processes features according to the length of the utterance by selectively merging the activations of different resolution branches in each block. Owing to these two components, our proposed model can extract speaker embeddings rich in time-spectral information and operate dynamically on length variations. Experimental results on the VoxCeleb1 test set consisting of various duration utterances demonstrate that RawNeXt achieves state-of-the-art performance compared to the recently proposed systems. Our code and trained model weights are available at https://github.com/wngh1187/RawNeXt., 5 pages, 2 figures, 4 tables, accepted to 2022 ICASSP as a conference paper

Various attention mechanisms are being widely applied to acoustic scene classification. However, we empirically found that the attention mechanism can excessively discard potentially valuable information, despite improving performance. We propose the attentive max feature map that combines two effective techniques, attention and a max feature map, to further elaborate the attention mechanism and mitigate the above-mentioned phenomenon. We also explore various joint training methods, including multi-task learning, that allocate additional abstract labels for each audio recording. Our proposed system demonstrates state-of-the-art performance for single systems on Subtask A of the DCASE 2020 challenge by applying the two proposed techniques using relatively fewer parameters. Furthermore, adopting the proposed attentive max feature map, our team placed fourth in the recent DCASE 2021 challenge., 5 pages, 2 figures, 5 tables, submitted to ICASSP 2022

This study aimed to evaluate the biomechanical properties in vitro and the bone regeneration of whitlockite (WH) compared with hydroxyapatite (HA) or β-tricalcium phosphate (β-TCP)-based material. We investigated the morphology and phase composition of the bone grafts using a scanning electron microscope and X-ray diffractometer patterns and tested the compressive strength. Four circular defects of 8 mm in diameter were created on the calvaria of twelve rabbits. One defect was left empty, and each of the other defects was filled with WH, HA, and β-TCP. At 4 and 8 weeks, the specimens were harvested to evaluate for the new bone formation and the remaining bone grafts. Regarding the biomechanical properties, the three grafts had a similar micropore size, and WH showed nanopores. The compressive strength of WH was higher than HA and β-TCP without statistical significance. The radiological and histomorphometric analyses demonstrated that the new bone formation was similar among the groups. The remaining bone graft of the WH group was greater than that of the HA and β-TCP groups at 4 weeks (p < 0.05), and the total bone area of the WH, HA, and β-TCP groups was greater than that of the other (p < 0.01). WH has excellent volumetric stability and osteoconductivity compared with HA and β-TCP.

Organisations have been increasingly aware of the importance of promoting environmental sustainability as one means of fulfilling a role in corporate social responsibility. Using data from 194 Hong Kong managerial and professional employees collected through two waves of online survey, we examined the impact of three proximal factors, autonomy at work, management involvement, and intrinsic motivation on workplace green behaviour (i.e., champion green behaviour, conventional green behaviour, paper use and electricity use). Our results showed that (1) employees’ intrinsic motivation and management expectations of acting green were significant predictors of workplace green behaviour; (2) autonomy at work was related only to champion green behaviour; and (3) the presence of committees or designated personnel on promoting green initiative was not related to employees’ workplace green behaviour.

Whitlockite (WH) is the second most abundant inorganic component of human bone, accounting for approximately 25% of bone tissue. This study investigated the role of WH in bone remodeling and formation in a mouse spinal fusion model. Specifically, morphology and composition analysis, tests of porosity and surface area, thermogravimetric analysis, an ion-release test, and a cell viability test were conducted to analyze the properties of bone substitutes. The MagOss group received WH, Group A received 100% beta-tricalcium phosphate (β-TCP), Group B received 100% hydroxyapatite (HAp), Group C received 30% HAp/70% β-TCP, and Group D received 60% HAp/40% β-TCP (n = 10 each). All mice were sacrificed 6 weeks after implantation, and micro-CT, hematoxylin and eosin (HE) staining, and Masson trichome (MT) staining and immunohistochemistry were performed. The MagOss group showed more homogeneous and smaller grains, and nanopores (