| Group |
2020-12 |
Status |
completed |
| Title |
Speech Enhancement Using Advanced Machine Learning Methods |
| Supervisor |
Wei-Ping Zhu and Mr. Mojtaba Hasannezhad |
| Description |
Nowadays, human-machine interfaces based on natural speech have become part of our daily life. Speech interfaces not only do facilitate human-machine interactions but also significantly enhance the efficiency and functionality of home automation, which is emerging as one of the most popular applications of internet of things. Many applications use automatic speech recognition (ASR) for interaction with intelligent electronic devices, such as Amazon Echo and Google Home. For these large-scale real-world applications, robustness to noise is an increasingly important issue since ASR needs to operate under difficult acoustic conditions. For instance, recognizing a command input to Amazon Echo in the living room involves dealing with a wide variety of noise sources, such as children’s voices, television or ambient music. Obtaining this clean command (speech) from the one degraded by surrounding noises is called speech enhancement (SE).
Recent studies in SE have resorted to machine learning as a primary tool to develop a data-driven method. Deep learning as a primary tool to develop a data-driven approach has marked a revolutionary advance in SE.
In this project, we investigate new SE methods using deep neural networks (DNN). The idea is to once train a DNN using noisy and clean speech, in the sense that we provide the noisy and clean speech as the input and output to DNN, respectively; and it automatically learns how to map the noisy speech to the clean one. In the testing stage, the noisy speech is fed to DNN and it outputs the clean speech.
|
| Student Requirement |
Students should have a basic knowledge of signal processing and machine learning. They should have basic Python coding skills. |
| Tools |
A laptop with python installed on it. |
| Number of Students |
3 or 4 stud |
| Students |
Alec Kurdjian
Alexia Chebeia
Cong-Vinh Vu
Gregory Cao
Kasra Laamerad |
| Comments: |
Email: weiping@ece.concordia.ca
Room: EV-5.225
Tel: (514) 848-2424 ext. 4132
|
| Links: |
|