An Optimized Recurrent Unit for Ultra-Low-Power Keyword Spotting

Low Resource Keyword Spotting

An Application of Recurrent Neural Networks to Discriminative Keyword Spotting

Keyword spotting is a detection task consisting in discovering the presence of specific spoken words in unconstrained speech. The majority of keyword spotting systems are based on generative hidden Markov models and lack discriminative capabilities. However, discriminative keyword spotting systems are based on the estimation of a posteriori probabilities at the frame-level, hence they make use ...

Convolutional Recurrent Neural Networks for Small-Footprint Keyword Spotting

Keyword spotting (KWS) constitutes a major component of human-technology interfaces. Maximizing the detection accuracy at a low false alarm (FA) rate, while minimizing the footprint size, latency and complexity are the goals for KWS. Towards achieving them, we study Convolutional Recurrent Neural Networks (CRNNs). Inspired by large-scale state-ofthe-art speech recognition systems, we combine th...

An Ultra-Low Power Variable Gain Current Mirror

A power efficient gain adjustment technique is described to realize programmable gain current mirror. The dissipation power changes over the wide gain range of structure are almost negligible. This property is in fact very interesting from power management perspective, especially in analog designs. The simple structure and constant frequency bandwidth are other ever-interesting merits of propos...

Cross-word sub-word units for low-resource keyword spotting

We investigate the use of sub-word lexical units for the detection of out-of-vocabulary (OOV) keywords in the keyword spotting task. Sub-word units based on morphological decomposition and character ngrams are compared. In particular, we examine the benefit of sub-word units that cross word boundaries. Experiments are performed on the IARPA Babel Turkish dataset. Our results demonstrate that cr...