Applied Speech and Audio Processing: With matlab examples
Psychoacoustic processing
Download 2.66 Mb. Pdf ko'rish
|
Applied Speech and Audio Processing With MATLAB Examples ( PDFDrive )
4.4. Psychoacoustic processing
75 does change the amplitude (firstly by allocating more signal power to the sound, and secondly by spreading the components into more critical bands). This relationship allows the ‘strengthening’ of a sound by the addition of related frequency components without causing undue concentration of energy at specific frequencies, as would be the case with direct amplification of those fundamental frequencies to achieve a perceptually similar result. This technique may allow a speech formant, obscured by localised noise, to be per- ceptually strengthened [23]. 4.4.3 Temporal masking release Post-stimulatory temporal masking (see Section 4.2.10), causing sounds occurring just after a loud tone or pulse to be perceived less loud, may be negated by detecting such events and applying an initial increase, followed by a decaying level of amplification after it. A knowledge of the exact amplitude at the ear of a listener is probably a necessary requirement for this processing. In a similar way for much louder sounds, allowances may be made for TTS (temporary threshold shift – see Section 4.2.6). 4.4.4 Masking and two-tone suppression If a wanted tone lies close to a masking tone of higher amplitude, it is known that the wanted tone will be masked (see Section 4.2.8). To alleviate this situation, the wanted tone could be amplified, strengthened by the addition of harmonics, or simply shifted slightly. If the wanted tone is harmonically related to other tones, we know that the shift may not be more than about 3% [4]. Techniques already exist for predicting the effects of masking for the purpose of enhancement, many of which are reported to perform fairly well [14,15,20,24,25]. 4.4.5 Use of correlated noise Remember that modulated noise, present in more than one critical band, allows the auditory system to use a correlation process to reduce its effects (refer to Section 4.2.9). In this way, we can conjecture that a frequency shifted, reduced amplitude version of that noise induced elsewhere in an unimportant part of the frequency spectrum should reduce its effect. 4.4.6 Binaural masking As noted in Section 4.2.13, judicious control of the signal fed to each ear of a listener may help to improve the perceived signal-to-noise level. For speech systems contaminated with additive white Gaussian noise, it appears that applying out-of-phase speech to each ear, whilst subject to the same noise field, can reduce the effect of the noise. |
Ma'lumotlar bazasi mualliflik huquqi bilan himoyalangan ©fayllar.org 2024
ma'muriyatiga murojaat qiling
ma'muriyatiga murojaat qiling