RESUMEN
Attention mechanism has been widely used in speech enhancement (SE) because, theoretically, it can effectively model the inherent connection of signal in time domain and spectrum domain. In this Letter, it is found that the attention over the entire frequency range hampers the inference for full-band SE and possibly leads to excessive residual noise and degradation of speech. To alleviate this problem, the local spectral attention is introduced into full-band SE model by limiting the span of attention. The ablation tests on three full-band SE models reveal that the local frequency attention can effectively improve overall performance.
RESUMEN
Objectives: Studies have established a link between periodontal disease and cardiovascular disease (CVD), but it is unclear whether there is a sex difference in their association. Methods: The PubMed, Embase, and Cochrane databases were searched until June, 21 2022. Cardiovascular outcomes included any CVD, myocardial infarction (MI), coronary heart disease (CHD), or stroke. Studies reported the prevalence of CVD in patients with periodontal disease and the relationship between periodontal disease and CVD. The study is registered with PROSPERO (CRD42022333663). The level of evidence and recommendations is assessed by the Grading of Recommendations for Assessment, Development and Evaluation (GRADE). Results: Twenty-six studies were included. In patients with periodontal disease, the prevalence of CVD was 7.2% [9 studies; 95% confidence interval (CI): 2.7-13.6%], and prevalence for CHD, hypertension, stroke, and heart failure was 6.6, 25.3, 1, and 1.1%, respectively. There was a significant association between periodontal disease and CVD in men [odds ratio (OR) = 1.22; 95% CI: 1.12-1.34] and women (OR = 1.11; 95% CI: 1.05-1.17), with no significant sex difference (P > 0.05). Conclusion: Cardiovascular disease is relatively common in patients with periodontal disease, and an increased risk of CVD is associated with periodontal disease independent of sex. Interventions targeting periodontal disease may be beneficial for CVD. Systematic review registration: https://www.crd.york.ac.uk/PROSPERO/, identifier CRD42022333663.
RESUMEN
The highly imbalanced power spectral density of full-band speech signals poses a significant challenge to full-band speech enhancement, and the commonly used spectral features that mimic the behavior of the human auditory system are not an optimal choice for full-band speech enhancement. In this paper, a learnable spectral dimension compression mapping is proposed to effectively compress the spectral feature along frequency, preserving high resolution in low frequencies while compressing information in high frequencies in a more flexible manner. Experimental results verify that the proposed method can be easily combined with different full-band speech enhancement models and achieve better performance.
