Human Sound Localization Research Articles

Individual differences in human sound localization behavior.

Results from studies of the apparent positions of virtual and real sound sources conducted over the past 5 years reveal large individual differences. To characterize these differences in global terms such as accuracy or variance does not tell the whole story. Listeners differ in the extent to which they make front–back or back–front confusions, the extent to which they report apparent positions above or below the horizontal plane, the resistance of their judgments of apparent position to interaural level imbalance, and other factors. It seems clear that some of these differences might be traced to differences in outer‐ear acoustics. Since pinna shapes vary tremendously from person to person, there are large individual differences in the resulting head‐related transfer functions (HRTFs). The cues provided by the direction‐dependency of HRTFs are known to be important, especially for determining apparent source elevation. HRTFs lacking in detail provide ineffective cues and thus listeners must adopt differing strategies for combining HRTF cues with other cues to determine source position. Unfortunately detailed analyses of both HRTF differences and differences in localization behavior does not reveal a simple relationship. [Work supported by NASA and NIDCD.]

Reassessment of the role of head movements in human sound localization

This project sought to quantify both the nature and consequences of a listener’s head movements during a sound localization task. The task required listeners to report the apparent direction and distance of virtual sound sources. The virtual source stimuli were synthesized from measurements of each listener’s own head-related transfer functions and were presented over headphones so that the acoustical consequences of head movements could be controlled. In one condition head movements were discouraged, and the stimuli delivered to the headphones were not influenced by movements of the listener’s head. In another condition the listener was encouraged to use small head movements to aid localization. The position of the listener’s head was tracked and the synthesis of each virtual source was modified in real time, in accordance with the head movements, to simulate a stationary external source. For those listeners who demonstrated a high rate of front–back confusions in the first condition, head movements apparently provided the information needed to resolve those confusions, since the confusion rate was near zero in the second condition. There was also some improvement in elevation perception with head movement information provided. Analyses of the head movement trajectories support the hypothesis that the major role of head movements is resolution of the confusions that result from reliance on interaural time and intensity cues. [Work supported by NIH, NASA, and the ONR.]

Neural network models of sound localization based on monaural information and based on binaural information

Neural networks were trained with back propagation to localize ‘‘virtual’’ sounds that could originate from any of 24 azimuths (−165° to +180°) and 6 elevations (−36° to +54°). The sounds (clicks) were filtered with head related transfer functions and transformed into 22-point quarter-octave spectra. The networks were composed of 22 to 44 input nodes, followed by 50 hidden nodes and 30 output nodes (24 azimuth nodes and 6 elevation nodes). With the ‘‘binaural’’ configuration, the interaural delay and the interaural difference spectrum were provided as inputs to the network. With the ‘‘monaural’’ configuration, separate networks were trained with the left-ear and the right-ear spectra; a third, arbitrator, network learned to localize based on the output of these two monaural networks (i.e., the activation levels of their azimuth and elevation nodes). The monaural configuration did not allow for binaural interaction in the traditional sense. Both configurations achieved performance comparable to humans, suggesting that for these conditions either monaural cues or binaural cues are sufficient to explain human sound localization. [Work supported by NIH-DC-00786, AFOSR-91-0289, and AFOSR-Task 2313V3.]

An artificial neural network model of human sound localization

An artificial neural network was trained to identify the location of sound sources using the head-related transfer functions (HRTFs) of Wightman and Kistler [J. Acoust. Soc. Am. 85, 858–867 (1989)]. The simulated signals were either filtered clicks or pure tones, with speaker placements separated in steps of 15 deg in azimuth of 18 deg in elevation. After the signals were passed through the HRTFs, the inputs to the nets were computed as the difference of left ear and right ear phase spectra or the difference of the power at the output of left and right ear third-octave or twelfth-octave filter banks. Back propagation was used to train the nets. Separate nets were trained for each signal type and for each type of input data. Better than 90% correct identification of the source speakers location can be achieved in either the horizontal or median planes. The results for the horizontal plane are compared to the predictions of the duplex theory of sound localization. [Work supported by AFOSR-91-0289 and AFOSR-Task 2313V3.]

Mismatch negativity to change in spatial location of an auditory stimulus

Auditory stimulus blocks were presented to 12 reading subjects. Each block consisted of 2 types, standard ( P = 90%) and deviant stimuli ( P = 10%), delivered in a random order. The only difference between these stimuli was their spatial location of origin. The subject always heard the standards as coming straight in front and the deviants from an angle of either 10, 45, or 90° to the right of the standards. The spatial locations were produced via earphones by introducing for low-frequency (600 Hz) tones an interaural phase difference and for high-frequency (3000 Hz) tones an interaural intensity difference. Standard and deviant stimuli were also delivered in more natural, free-field, conditions via differently positioned loudspeakers. The deviant tones elicited an event-related brain potential component called the mismatch negativity (MMN), followed by a P3a component. Thus changes in spatial location of an auditory stimulus produced by following either one of the two main principles of human sound localization elicited the MMN. Consequently, it was concluded that the spatial location of a sound source is coded in the hypothesized neuronal stimulus traces reflected by the MMN and, further, that a change in this location is automatically detected by the brain by means of the MMN generator process.

Review of Spatial Hearing: The Psychophysics of Human Sound Localization.

Review of Spatial Hearing: The Psychophysics of Human Sound Localization.

Model of human sound localization based on spectral pattern recognition

In the course of our research on sound localization we have measured direction‐dependent pinna transfer functions of eight observers. The magnitudes of these functions have peaks and valleys that vary in spectral locus across subject and across source location. We addressed the question of whether these spectral patterns contain sufficient information for accurate localization of sounds in space. We used standard multivariate statistical techniques to develop a model of sound localization based on pattern recognition principles. A Bayesian classification procedure is employed to make a decision as to the most probable source location. The errors made by this procedure are compared to the errors made by subjects when judging source location in free‐field listening conditions. Preliminary results are extremely encouraging. The source positions for which the model makes errors match quite well those for which the subjects make errors. [Work supported by NSF and NIH.]

Acuity of the human sound localization response underwater.

Minimum Audible Angles (MMA) for divers were determined for 3.5 kHz, 6.5 kHz, and broad-band white-noise pulses having a duration of 100 msec, a rise-decay time of 10 sec, and a repetition rate of 1/sec. In Expt. 1 six divers performed the right-left discrimination with no training or feedback and the mean MAAs obtained were 21.5° for the 3.5-kHz signal, 14.5° for the 6.5-kHz signal, and 9.8° for the white-noise signal. In Expt. 2 four of the original six divers were given feedback from their performance with the white-noise signal at a single small angle. The mean MAAs obtained after training were 11.3° at 3.5 kHz, 11.5° at 6.5 kHz, and 7.3° with white noise. It was concluded that if one assumes that the MAA reflects the general level at which the binaural system operates then it seems likely that man will be an effective sound navigator in the water.

Underwater Sound Localization: Preliminary Information

A series of studies has been initiated investigating the nature of human sound localization underwater; the research is based in part on the discussions and work of Sivian, Bauer, Torick, and others. The first step in the project was the development of Diver Auditory Localization System (DALS), which consisted of an underwater PVC structure with a series of five 8-ft arms beneath which was slung a subject's chair. DALS allows the diver/subject's head either to be held in a rigidly fixed position or to be moved, with his body fixed. The five arms allow placement—at ear level—of projectors at angles, to the diver, of 0°, 45°, 90°, 270°, and 315°. In this first experiment, six diver/talkers were positioned in DALS, at a depth of 50 ft, with their bodies fixed but their heads mobile; four different experimental signals 250-, 1000-, 6000-Hz sinusoids and thermal noise) were transduced via J-9 projectors. The site of the project was NRL's Underwater Sound Reference Division's Bugg Spring facility. Each signal was presented to the listeners five times randomly from each of the five transducers; each presentation was in the nature of five 1-sec bursts 40 dB re hearing threshold. Listeners responded by means of a specially constructed five-position underwater switch coupled to an IBM keypunch at the surface. Data show unexpected localization ability by subjects (theory would predict no such ability); the mean correct scores for the six subjects were: (1) 250 Hz: 52.7% (2) 1 kHz: 37.3% (3) 6 kHz: 24.0%; and (4) noise: 57.3%. Thus, only the data for 6 kHz approached chance, and it can be concluded that the subjects in this experiment exhibited greater skills at underwater sound localization than was expected. Further studies presently are in progress. [This project was supported by the Physiological Psychology Branch of the Office of Naval Research.]