![]() |
||||||||
SRoL Sitemap
First page |
The background noise analysis The background noise is a random signal, seldom stationary, mixed with some deterministic perturbations of network frequency perturbations type. To ensure the quality of the recordings, the noise amplitude must be much smaller than the signal amplitudes. The recordings have been effectuated in a laboratory with reduced noise, but we have not disposed of a completely phonically isolated room. A good quality recording has a noise in whose spectrum the value -80dB is not surpassed for neither spectral component, while the F1 formant has a majored level with at least 30dB, and the amplitudes for the superior formants are typically with more then 15-20dB over the noise level. The present recordings do not satisfy completely those conditions, being just informative. Picking the microphone The microphone is the first element in the sounds conversion and has an essential impact on the recordings quality. A limited tape microphone, of reduced sensibility or with high noise can compromise the recordings.An above-grade microphone has reduced noise and a good ratio of signal/noise. The omni-directional feature allows for a high sensibility to the environment noises and this is not desired. The sounds have been recorded using a SONIC Stereo Dynamic Headphones HP-259 with the following features:
A special attention must be given to the microphone, because noises and distortions appear due to deficient placement. Holding the microphone to close to the mouth can lead to the amplifier saturation effect, having as result strong signal distortion. It is recommended to place the microphone under the mouth, approximately at the chin level, a few centimeters from it. The distance from the chin must be equal with the distance to the lips. The soundboard and the corresponding drivers Most recent computers have motherboards containing incorporated soundboards for processing signals from the microphone and generating audio signal for speakers and headphones (the quality of these soundboards differs substantially for each model). The motherboard of the computer on which the recordings were made is MB FOXCONN 760 GXK8MC-S, with incorporated Sound MAX Digital Audio sound card made by Analog Devices, with the following characteristics:
The recordings were performed using GoldwaveTM 5.0 at a sampling frequency of 22050 Hz with PCM signed (24 bits mono). The subjects were informed prior to recording of the project's objectives and assured of the confidentiality of the personal information. The subjects signed an informed consent in accordance with the US Food and Drug Administration's Protocol of Protection of Human Subjects and the Ethic Principles of the American Acoustic Association regarding Research Involving Human Subjects . Methodology for the construction and analysis of corpus SRoL (in Romanian) © 2010 Horia-Nicolai Teodorescu |
|||||||
This page was last updated by H.-N. Teodorescu and M. Feraru on Copyright©2006 - SRoL |