Speech Recognition: Keyword Spotting Through Image Recognition
Authors:
Sanjay Krishna Gouda,
Salil Kanetkar,
David Harrison,
Manfred K Warmuth
Abstract:
The problem of identifying voice commands has always been a challenge due to the presence of noise and variability in speed, pitch, etc. We will compare the efficacies of several neural network architectures for the speech recognition problem. In particular, we will build a model to determine whether a one second audio clip contains a particular word (out of a set of 10), an unknown word, or silen…
▽ More
The problem of identifying voice commands has always been a challenge due to the presence of noise and variability in speed, pitch, etc. We will compare the efficacies of several neural network architectures for the speech recognition problem. In particular, we will build a model to determine whether a one second audio clip contains a particular word (out of a set of 10), an unknown word, or silence. The models to be implemented are a CNN recommended by the Tensorflow Speech Recognition tutorial, a low-latency CNN, and an adversarially trained CNN. The result is a demonstration of how to convert a problem in audio recognition to the better-studied domain of image classification, where the powerful techniques of convolutional neural networks are fully developed. Additionally, we demonstrate the applicability of the technique of Virtual Adversarial Training (VAT) to this problem domain, functioning as a powerful regularizer with promising potential future applications.
△ Less
Submitted 24 November, 2020; v1 submitted 10 March, 2018;
originally announced March 2018.
Unusual structures inherent in point pattern data predict colon cancer patient survival
Authors:
Charlotte M. Jones-Todd,
Peter Caie,
Janine Illian,
Ben C. Stevenson,
Anne Savage,
David J. Harrison,
James L. Bown
Abstract:
Cancer patient diagnosis and prognosis is informed by assessment of morphological properties observed in patient tissue. Pathologists normally carry out this assessment, yet advances in computational image analysis provide opportunities for quantitative assessment of tissue. A key aspect of that quantitative assessment is the development of algorithms able to link image data to patient survival. H…
▽ More
Cancer patient diagnosis and prognosis is informed by assessment of morphological properties observed in patient tissue. Pathologists normally carry out this assessment, yet advances in computational image analysis provide opportunities for quantitative assessment of tissue. A key aspect of that quantitative assessment is the development of algorithms able to link image data to patient survival. Here, we develop a point process methodology able to describe patterns in cell distribution within cancerous tissue samples. In particular, we consider the Palm intensities of two Neyman Scott point processes, and a void process developed herein to reflect the spatial patterning of the cells. An approximate-likelihood technique is taken in order to fit point process models to patient data and the predictive performance of each model is determined. We demonstrate that based solely on the spatial arrangement of cells we are able to predict patient survival.
△ Less
Submitted 16 May, 2017;
originally announced May 2017.
