Performance of Oversampled Polyphase Filterbank Inversion via Fourier Transform

Frequency channelization is a fundamental signal processing operation employed across various domains, including communications and radio astronomy. The polyphase filterbank (PFB) represents an efficient and versatile means of channelization. When strict constraints are placed on the allowable spectral leakage between neighboring channels, an oversampled PFB design is advantageous. A helpful consequence of the oversampling is that inversion of the PFB to recover high temporal resolution is simplified and can be accomplished accurately using Fourier transforms. We describe this inversion approach and identify key design considerations. We examine the residual error and spectral/temporal leakage behavior when a channelizer and its corresponding inverter are cascaded, concluding that near-perfect reconstruction can be approached with appropriate selection of PFB and inverter design parameters.

Segmentation-Enhanced CycleGAN

Algorithmic reconstruction of neurons from volume electron microscopy data traditionally requires training machine learning models on dataset-specific ground truth annotations that are expensive and tedious to acquire. We enhanced the training procedure of an unsupervised image-to-image translation method with additional components derived from an automated neuron segmentation approach. We show that this method, Segmentation-Enhanced CycleGAN (SECGAN), enables near perfect reconstruction accuracy on a benchmark connectomics segmentation dataset despite operating in a “zero-shot” setting in which the segmentation model was trained using only volumetric labels from a different dataset and imaging method. By reducing or eliminating the need for novel ground truth annotations, SECGANs alleviate one of the main practical burdens involved in pursuing automated reconstruction of volume electron microscopy data.