Theability to localize sound provides a great deal of benefits to humans as wellas animals.
Localization allows us to be aware of dangerous situations, itallows for orientation of visual attention, it aids in the discrimination ofsounds from complex soundscapes such as distinguishing between differentinstruments in an orchestra, as well as aids in speech in noise perception.Human listeners are able to localize sounds in space in terms of assigningdirection and distance to the perceived auditory image. This is essential inspatial hearing which relies on physical directional acoustic features (i.e.azimuth, distance, and elevation) which are the consequence of acousticfiltering of the sound by the pinna, head, and torso.
These filters aregenerated from the sound pathway (sound source to sound receiver which isusually placed at the entrance of the ear canal).Thereare various localization cues such as interaural difference in time (ITD) andsound level (ILD) as well as the measurement of the minimum audible angel (MAA).The MAA is defined as the smallest detectable angular separation between twosound sources relative to the head (i.e. azimuthal differences). Acousticfiltering can be described by head related transfer functions (HRTFs). HRTFsare used for spatial hearing, fitting hearing aids, and presenting virtual binauralaudio signals via headphones in virtual auditory displays.
There are differenttypes of HRTFs such as listener-specific HRTFs and generic HRTFs.Listener-specific HRTFs are acoustically measured by placing a small microphoneat the entrance of the listener’s ear canal. Listener-specific HRTFs are moreaccurate for fitting children with hearing aids as well as providing accuratesound localization performance via headphones; however, this type of HRTF isgenerally time-consuming to obtain due to the individual specific HRTF thatneeds to be obtained. Generic HRTFs are obtained through mannequinsrepresenting an average of the human population and is more time efficient. Ina sound system, there is an input sound which is in the form of an impulse,filters in which the sound travels through, and the output sound that is alsocalled the impulse response.
Filters either result in sound gain orattenuation. The impulse response is another method of determining localizationand with what impulse generates a greater magnitude in impulse response. Agreater magnitude of response means that at this impulse, localization isgreater. This means that the MAA would be smaller and the individual has astrong ability to locate a sound source with an impulse that generates agreater response. Literature ReviewInAlgazi and Duda’s (2002) study, they looked at approximating HRTF usinggeometric models of the head and torso.
Specifically, they analyzed lowfrequency behavior at low elevations. They looked into the torso reflectionsand torso shadow and how these provide significant elevation cues at a range ofelevations. They compared their HRTF model with acoustic measurements in separateplanes (horizontal, median, and frontal) that verified the essential validityof the estimates of the HRTF for the KEMAR mannequin which will later beexplained in this study. InAlagazi, Avendano, and Duda’s (2000) study, they looked at elevationlocalization and HRTF analysis at low frequencies. In this study, elevationangles around four cones of confusion were reported by six test subjects. Therewas a reduction in performance when compared to baseline test and was afunction of azimuth and was highest in the median plane.
Subjects were able to approximateelevation when the loudspeaker was positioned far from the median plane. Analysisof HRTFs in this study established the presence of elevation dependent featuresat low frequencies. InIida, Ishii, Nishioka’s (2014) study, they looked at the personalization ofHRTFs in the median plane based on the anthropometry of the listener’s pinnae. Thisstudy’s test method utilized localization tests in the upper median plane onfour test subjects.
The results revealed that the best-matching HRTFs provided roughlythe same performance as the listener’s own HRTFs for the target directions ofthe front and rear for all four test subjects decreased. The presentation of thelocalization for particular subjects reduced for the upper target direction. InGamper’s (2013) study, he looked at HRTF interpolation in azimuth, elevation,and distance. This study looks at the lack of algorithms and tools to make useof virtual sounds in the near-field. The data is represented in a 3-dimensionalgraph that displays azimuth, distance, and elevation. This visual display showsthe three measures and the weight each measurement holds on HRTFs. InHarder, Paulsen, Larsen, Laugesen, Mihocic, and Majdak’s (2016) study, theylooked at the framework for geometry acquisition, simulation and measurement ofHRTFs with a focus on hearing-assistive devices (HADs).
This is due todiscovering HRTFs essential in fitting HADs to provide accurate soundlocalization performance. They found that the ability to provide localizationcues to HAD users is a possibility. PurposeThe purpose ofthis study was to compare the magnitude of head related impulse responsechanges at 0, 90, and 180 degree azimuths.
These changes are important forunderstanding which azimuth provides the best circumstance for localization ofsound sources. Methods MaterialsThematerials used in this study consisted of a loudspeaker that played whitenoise. This white noise acted as the input in the sound system that wasproduced in this study. The Knowles Electronics Mannequin for Acoustic Research(KEMAR) was used to measure the output in the form of head related impulseresponses via a small microphone inside both ear canals of KEMAR as seen in figure1. KEMAR is the first head and torso simulator designed especially for acousticresearch. The head and torso size of KEMAR is based on the average of 5,000 maleand female adult dimensions. Realistic measurement in a sound system experiencedby a human is achievable due to the design of KEMAR.
The sound system in thisstudy looked at an input sound being produced by a loudspeaker. This soundpasses through various filters with the main filter that will be discussed inthis study being the head related transfer function (HRTF) as mentionedpreviously. Then an output sound is produced and recorded by KEMAR. A MATLAB algorithmwas used to analyze the data recorded by KEMAR and displayed in the form ofgraphs represented with the head related impulse response in decibel as afunction of frequency in hertz.
The head related impulse responses wererecorded as an average of the left and right ear as the sound source wasplayed.