EXPERIMENTS ON TRAINING THE SOUND LOCALIZATION ABILITIES: A SYSTEMATIC REVIEW

Abstract

Hearing is an extraordinary sense, as the audio signals perceived at the eardrum level of both ears build and define the entire surrounding environment. Sound is seized from all directions, and the listener can identify the distance of the sound source, the amplitude, the arrival time at each ear and the variations of amplitude that differ with changing frequencies. 3D sound can be an efficient sensory substitution method that models the elements of the physical world into relevant information for the visually-impaired people. 3D sound is generated by using spectral cues that define how a sound is perceived at a specific location. The majority of virtual-reality audio devices are based on the Head Related Transfer Function (HRTF), a mathematical transformation that describes how the sound is being modified en route from the source until it reaches the inner ear. This article analyzes several experiments performed in order to investigate human sound localization abilities involved a training and a test session. To test the users' capability to distinguish the sound sources located at various positions in space (azimuth- in the horizontal plane and elevation- in the vertical plane), the sound was played once or several times, as the acoustic signal varied in shape (pink noise, white noise, harmonic sine wave), frequency and duration. Some conclusion indicate that best spatial recognition is achieved by using the white pulse Gaussian noise with a duration of around 100-300ms. Pink and white noise are the most separable sounds, while sine waves are annoying and difficult to localize in the horizontal plane. In addition to this, azimuth spatial resolution is superior to elevation spatial resolution because of the stereoscopic properties of the human auditory system. Also, the sound resolution is better when the initialization phase (training session when the user listens to the sounds for several minutes) lasts longer and when the acoustic signal is periodic and of long continuous duration. However, we recognized that there are important limitation of the experiments, that failed to take into account and investigate important aspects like sound duration, continuity and effects and also the use of structured performance feedback in the training process. We present, in this paper, brief arguments why we believe that such aspects are worth of further investigation and can lead to substantial improvements.