Projects:2020s2-7410 Speech Enhancement for Automatic Speech Recognition

An increasing number of applications require the joint use of signal processing and AI techniques on time series and sensor data. These techniques can be used for the reduction of noises such as air conditioning, computer fan, or environmentally generated noises such as in a street, an airport, in a metro station, or in an airplane cockpit. Developing AI models for signal obtained from a variety of situations as exemplified above is not trivial, but these have been attempted using Recurrent and Convolutional Networks such as Speech Enhanced Generative Adversarial Neural networks (SEGAN).

Introduction

Project Aims

Use HARK with PyKALDI on the High-Performance Computer.

Develop an algorithm using HARK for noise processing on HPC.

Evaluate the performance of HARK relative to a number of noise types.

Perform speaker identification using PyKALDI on HPC.

Project Team

Students

• Muhammad Haniff Derani
• Shuyang Shen

Supervisors

• Dr. Said Al-Sarawi (
• Dr. Ahmad Hashemi-Sakhtsari (DST Group)

Abstract