Speech signals exhibit strong temporal dependencies and local prosodic patterns that are crucial for accurate depression detection. However, existing methods often fail to adequately model the long-range temporal dependencies of speech signals, and predominantly focus on single-granularity features, thus limiting their capacity to represent non-stationary and intermittent pathological patterns. To address these limitations, we propose DB-SMGA, a novel dual-branch sequential multi-granularity attention model. The model processes former and latter subsequences separately through a dual-branch architecture to capture the long-range temporal evolution of the entire sequence. Each branch contains the time-frequency separation (TFS) module and the hybrid dense-sparse attention (HDSA) module, which together integrate transient and persistent depressive features. Furthermore, through hierarchical multi-granularity feature fusion, DB-SMGA achieves complementary enhancement of sequential features across both branches. Experimental results on two public datasets demonstrate that DB-SMGA significantly outperforms mainstream models, further highlighting its superiority in capturing the sequential evolution patterns of speech signals.

Rotating machinery prognostics plays an important role in promoting reliability and efficiency in the operation of machinery and reducing maintenance costs. This paper proposes a novel bearing remaining useful life prediction approach which puts emphasis on deriving effective features from raw vibration data. To enhance the degradation feature extraction, a non-recursive method named variational mode decomposition (VMD) is adopted to decompose the raw vibration data into several principal modes, then feature smoothing with a local regression filter is performed and some suitable features are selected by evaluating feature fitness using monotonicity and correlation analysis. Based on the selected features, long short-term memory (LSTM) network is introduced for bearings RUL prediction. Numerical experiments with real bearing dataset exhibit the effectiveness and superiority of the proposed approach in comparison to other data-driven approaches.

SUMMARY: Technological advances in high-throughput sequencing necessitate improved computational tools for processing and analyzing large-scale datasets in a systematic automated manner. For that purpose, we have developed PRADA (Pipeline for RNA-Sequencing Data Analysis), a flexible, modular and highly scalable software platform that provides many different types of information available by multifaceted analysis starting from raw paired-end RNA-seq data: gene expression levels, quality metrics, detection of unsupervised and supervised fusion transcripts, detection of intragenic fusion variants, homology scores and fusion frame classification. PRADA uses a dual-mapping strategy that increases sensitivity and refines the analytical endpoints. PRADA has been used extensively and successfully in the glioblastoma and renal clear cell projects of The Cancer Genome Atlas program. AVAILABILITY AND IMPLEMENTATION: http://sourceforge.net/projects/prada/ CONTACT: [email protected] or [email protected] SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.