Real-Time Multiple Audio Beamforming System
(2018)Department of Automatic Control
- Abstract
- The Cocktail party problem is a known problem within audiology and relates to a person’s ability to understand and separate speech from a noisy background with multiple speakers. This master’s thesis, in collaboration with Eriksholm Research Center, part of the company Oticon, developed a real-time system to help people with Hearing Impairment (HI) handle this problem by using a microphone array and a Minimum Variance Distortionless Response (MVDR) beamformer. The resulting beams yield an Articulation Index Weighted Directivity Index (AI-DI) of above 15 dB. Instead of using a single beam and removing all ambient sources, the system will let the user keep their situational awareness by using Head- Related Transfer Functions (HRTFs) and... (More)
- The Cocktail party problem is a known problem within audiology and relates to a person’s ability to understand and separate speech from a noisy background with multiple speakers. This master’s thesis, in collaboration with Eriksholm Research Center, part of the company Oticon, developed a real-time system to help people with Hearing Impairment (HI) handle this problem by using a microphone array and a Minimum Variance Distortionless Response (MVDR) beamformer. The resulting beams yield an Articulation Index Weighted Directivity Index (AI-DI) of above 15 dB. Instead of using a single beam and removing all ambient sources, the system will let the user keep their situational awareness by using Head- Related Transfer Functions (HRTFs) and directing beams at multiple sources and merging the result with an emphasis on the desired source. With the implemented eye-steering technique, the user has the ability to choose a desired source by looking at it. Based on Short-time Objective Intelligibility (STOI) measurements, the system more than doubles the estimated speech intelligibility in situations like the Cocktail party problem. Proposed techniques for implementations of Sound Source Localization (SSL) and Simultaneous Localization and Mapping (SLAM) are included in the report, which would let the system track sources in real-time. (Less)
Please use this url to cite or link to this publication:
http://lup.lub.lu.se/student-papers/record/8953835
- author
- Lindqvist, Johan and Sollenberg, Martin
- supervisor
- organization
- year
- 2018
- type
- H3 - Professional qualifications (4 Years - )
- subject
- report number
- TFRT-6055
- ISSN
- 0280-5316
- language
- English
- id
- 8953835
- date added to LUP
- 2018-06-29 12:14:21
- date last changed
- 2018-06-29 12:14:21
@misc{8953835,
abstract = {{The Cocktail party problem is a known problem within audiology and relates to a person’s ability to understand and separate speech from a noisy background with multiple speakers. This master’s thesis, in collaboration with Eriksholm Research Center, part of the company Oticon, developed a real-time system to help people with Hearing Impairment (HI) handle this problem by using a microphone array and a Minimum Variance Distortionless Response (MVDR) beamformer. The resulting beams yield an Articulation Index Weighted Directivity Index (AI-DI) of above 15 dB. Instead of using a single beam and removing all ambient sources, the system will let the user keep their situational awareness by using Head- Related Transfer Functions (HRTFs) and directing beams at multiple sources and merging the result with an emphasis on the desired source. With the implemented eye-steering technique, the user has the ability to choose a desired source by looking at it. Based on Short-time Objective Intelligibility (STOI) measurements, the system more than doubles the estimated speech intelligibility in situations like the Cocktail party problem. Proposed techniques for implementations of Sound Source Localization (SSL) and Simultaneous Localization and Mapping (SLAM) are included in the report, which would let the system track sources in real-time.}},
author = {{Lindqvist, Johan and Sollenberg, Martin}},
issn = {{0280-5316}},
language = {{eng}},
note = {{Student Paper}},
title = {{Real-Time Multiple Audio Beamforming System}},
year = {{2018}},
}