This year I am looking for a small team of MSc students to work on a set of four very closely related but distinct projects in the exciting new field of 'active hearing'. The students will work together during the 1st semester and will then take different research directions over the summer.
- JPB-MSc-1: Active hearing: sound source direction estimation with one ear (Isma Zulfiqar)
- JPB-MSc-2: Active hearing: sound source distance estimation with one and two ears (Wadha AlMattar)
- JPB-MSc-3: Active hearing: sound source distance estimation with one and two ears (Xiangtao Ma)
JPB-MSc-1: Active hearing: sound source direction estimation with one ear
Description
Humans and other animals are very good at using their ears to localise sound sources. Scientific accounts of sound source localisation tend to concentrate on the ability to determine the direction of the sound, i.e. whether it is coming from the left or right. According to theory, the direction is estimated by comparing the amplitude and arrival time of the sound in the left and the right ear. However, this simple account ignores the fact that animals move their heads around in a strategic manner while they are listening.
Listening with ears that are making controlled movements – active listening – is a big advantage over listening with stationary ears. By moving the ears it is possible to estimate the direction far more reliably. In fact, the precise 3D position of the sound can be computed. In some conditions, sounds can be localised with just a single (moving) ear.
Active hearing is a very new research area meaning that genuinely interesting and novel work can be done with little resource. This project will start by investigating how humans behave. The listeners will be given a series of challenging sound localisation tasks to perform. During the tasks their heads will be held fixed or allowed to move freely. We will use special microphones that sit inside the listener’s ears to record the signals they hear. A sophisticated head tracking device will accurately record their head movements. Using this data we will try to design automatic hearing systems that are able to perform the same task that the humans are trying to perform. (The ultimate aim is to learn smart ways to control a listening robot.)
The project has been subdivided into two sound source estimation tasks – sound distance and sound direction – and each of these tasks can be studied when listening with two ears (binaural) or with one ear (monaural). Students will work together to perform the recordings and the post processing of the recorded data. Later they will be assigned to one task (angle/distance) and one condition (monaural/binaural). Stationary ear hearing techniques will be compared with techniques that are able to exploit head motion.
Requirements
- A willingness to undertake practical work.
- The ability to work as part of a small team.
- The confidence to do some real research.
- An interest in using computers to better understand how humans work!
- Programming skills.
Initial reading
- Cooke, Lu, Lu and Horaud, “Active hearing, active speaking”, In: Auditory signal processing in hearing-impaired listeners (eds. T. Dau, J.M. Buchholz, J.M. Harte, T.U. Christiansen). 2008 pdf
- Research papers on sound source localisation and distance estimation (contact me)
JPB-MSc-2: Active hearing: sound source distance estimation with one and two ears
Description
Humans and other animals are very good at using their ears to localise sound sources. Scientific accounts of sound source localisation tend to concentrate on the ability to determine the direction of the sound, i.e. whether it is coming from the left or right. According to theory, the direction is estimated by comparing the amplitude and arrival time of the sound in the left and the right ear. However, this simple account ignores the fact that animals move their heads around in a strategic manner while they are listening.
Listening with ears that are making controlled movements – active listening – is a big advantage over listening with stationary ears. By moving the ears it is possible to estimate the direction far more reliably. In fact, the precise 3D position of the sound can be computed. In some conditions, sounds can be localised with just a single (moving) ear.
Active hearing is a very new research area meaning that genuinely interesting and novel work can be done with little resource. This project will start by investigating how humans behave. The listeners will be given a series of challenging sound localisation tasks to perform. During the tasks their heads will be held fixed or allowed to move freely. We will use special microphones that sit inside the listener’s ears to record the signals they hear. A sophisticated head tracking device will accurately record their head movements. Using this data we will try to design automatic hearing systems that are able to perform the same task that the humans are trying to perform. (The ultimate aim is to learn smart ways to control a listening robot.)
The project has been subdivided into two sound source estimation tasks – sound distance and sound direction – and each of these tasks can be studied when listening with two ears (binaural) or with one ear (monaural). Students will work together to perform the recordings and the post processing of the recorded data. Later they will be assigned to one task (angle/distance) and one condition (monaural/binaural). Stationary ear hearing techniques will be compared with techniques that are able to exploit head motion.
Requirements
- A willingness to undertake practical work.
- The ability to work as part of a small team.
- The confidence to do some real research.
- An interest in using computers to better understand how humans work!
- Programming skills.
Initial reading
- Cooke, Lu, Lu and Horaud, “Active hearing, active speaking”, In: Auditory signal processing in hearing-impaired listeners (eds. T. Dau, J.M. Buchholz, J.M. Harte, T.U. Christiansen). 2008 pdf
- Research papers on sound source localisation and distance estimation (contact me)
JPB-MSc-3: Active hearing: sound source distance estimation with one and two ears
Description
Humans and other animals are very good at using their ears to localise sound sources. Scientific accounts of sound source localisation tend to concentrate on the ability to determine the direction of the sound, i.e. whether it is coming from the left or right. According to theory, the direction is estimated by comparing the amplitude and arrival time of the sound in the left and the right ear. However, this simple account ignores the fact that animals move their heads around in a strategic manner while they are listening.
Listening with ears that are making controlled movements – active listening – is a big advantage over listening with stationary ears. By moving the ears it is possible to estimate the direction far more reliably. In fact, the precise 3D position of the sound can be computed. In some conditions, sounds can be localised with just a single (moving) ear.
Active hearing is a very new research area meaning that genuinely interesting and novel work can be done with little resource. This project will start by investigating how humans behave. The listeners will be given a series of challenging sound localisation tasks to perform. During the tasks their heads will be held fixed or allowed to move freely. We will use special microphones that sit inside the listener’s ears to record the signals they hear. A sophisticated head tracking device will accurately record their head movements. Using this data we will try to design automatic hearing systems that are able to perform the same task that the humans are trying to perform. (The ultimate aim is to learn smart ways to control a listening robot.)
The project has been subdivided into two sound source estimation tasks – sound distance and sound direction – and each of these tasks can be studied when listening with two ears (binaural) or with one ear (monaural). Students will work together to perform the recordings and the post processing of the recorded data. Later they will be assigned to one task (angle/distance) and one condition (monaural/binaural). Stationary ear hearing techniques will be compared with techniques that are able to exploit head motion.
Requirements
- A willingness to undertake practical work.
- The ability to work as part of a small team.
- The confidence to do some real research.
- An interest in using computers to better understand how humans work!
- Programming skills.
Initial reading
- Cooke, Lu, Lu and Horaud, “Active hearing, active speaking”, In: Auditory signal processing in hearing-impaired listeners (eds. T. Dau, J.M. Buchholz, J.M. Harte, T.U. Christiansen). 2008 pdf
- Research papers on sound source localisation and distance estimation (contact me)