Ambient sound reproduction through inter-aural earphones

• Main Authors: Cheng, Samuel Charles U., Joaquin, Patrick Bernard L., Ong, Charles Daniel S., Tiu, Abigail Beatrice E.
• Format: text
• Language: English
• Published: Animo Repository 2012
• Online Access: https://animorepository.dlsu.edu.ph/etd_honors/348
• Institution: De La Salle University

The problem with the design of the canalphones, although it lessens the risk of users to be aurally impaired by decreasing the tendency of constant volume changes, is that it decreases the awareness of the user by greatly attenuating important spatial auditory cues. This attenuation may lead to accidents and in some cases, work being done simultaneously while using the earphones may not be attended to properly. As a solution to the given problem, this research proposes an accurate and reproduction by introducing the necessary sound characteristics that were attenuated by the canalphones. The system records the sound, process the recorded file, and plays the processed signal in a pair of canalphones. Accurate sound reproduction is achieved through the concepts of Head-Related Transfer Functions (HRTF) and Head-Related Impulse Response (HRIR). A microphone placement that uses two omnidirectional was chosen by performing several tests. Based on the results, two microphones placed at a fixed distance above the ear is the most appropriate position. The system is able to capture sounds coming from any direction and is able to manipulate it such that the sound, when played back, provided with the necessary cues to approximately identify the original sound source location. Three methods (Inter-aural Level Difference Exaggeration, Batteau Pinna Model and Direction Specific Pinna model) were compared to find the method that performs better for locations of both azimuth and distance. After several tests, the Direction Specific Pinna model is chosen as the method for azimuth localization. From the results, the subjects were able to localize sound sources coming from front, front left, and front-right with 90%, 75% and 70% of the total test files localized correctly, compared to 35%, 25% and 45% using the Batteu Pinna model coming from the same azimuth both with a human voice as a stimulus.