3 accepted papers
Detecting emotions directly from a speech signal plays an important role in effective human-computer interactions. Existing speech emotion recognition models require massive computational and storage resources, making them hard to implement concurrently with other machine-interactive tasks in embedd…
In this paper, a novel multi-head multi-layer perceptron (MLP) structure is presented for implicit neural representation (INR). Since conventional rectified linear unit (ReLU) networks are shown to exhibit spectral bias towards learning low-frequency features of the signal, we aim at mitigating this…
In this paper, we study the realization of any given fully-digital precoder (FDP) by hybrid analog/digital precoding (HADP) in wide-band mmWave systems. We first formulate the massive-MIMO OFDM-based HADP system design and then, introduce the notion of perfect reconstruction at sampling point (PRSP)…