Meta-learning based Selective Fixed-filter Active Noise Control System with ResNet Classifier

The selective fixed-filter strategy is popular in industrial applications involving active noise control (ANC) technology, which circumvents the time-consuming online learning process by selecting the best-matched pre-trained control filter. However, the existing selective fixed-filter ANC (SFANC) based algorithms classify noises in frequency band, which is not a reasonable approach. Moreover, they pre-train the control filter utilizing only a single noise segment, leading to inaccurate estimation and undesirable noise cancellation performance when dealing with dynamically time-varying noise. Inspired by the applicability of meta-learning to various models utilizing gradient descent technique, this paper proposes a novel meta-learning based SFANC system, wherein the fixed-filters that may not be optimal for specific types of noises but can rapidly adapt to previously unseen noise conditions are pre-trained. To address the mismatch issue between meta-learning update methods and ANC requirements while enhancing the receptive field and convergence speed of control filters, a multiple-input batch processing strategy is utilized in pre-training. Simulations based on the common ESC-50 noise dataset are performed and demonstrate the superiorities of the proposed method in terms of classification accuracy, convergence speed, and steady-state noise cancellation.

Automated Measurements of Key Morphological Features of Human Embryos for IVF

A major challenge in clinical In-Vitro Fertilization (IVF) is selecting the highest quality embryo to transfer to the patient in the hopes of achieving a pregnancy. Time-lapse microscopy provides clinicians with a wealth of information for selecting embryos. However, the resulting movies of embryos are currently analyzed manually, which is time consuming and subjective. Here, we automate feature extraction of time-lapse microscopy of human embryos with a machine-learning pipeline of five convolutional neural networks (CNNs). Our pipeline consists of (1) semantic segmentation of the regions of the embryo, (2) regression predictions of fragment severity, (3) classification of the developmental stage, and object instance segmentation of (4) cells and (5) pronuclei. Our approach greatly speeds up the measurement of quantitative, biologically relevant features that may aid in embryo selection.