Washington, District of Columbia
June 23, 1996
June 23, 1996
June 26, 1996
1.55.1 - 1.55.7
ALTERATION OF THE FREQUENCY PROFILE OF VOICE SIGNALS
Essaid Bouktache, Chandra R Sekhar, Omer Farook Purdue University Calumet
Abstract: This paper explores the feasibility of a fi-equency translation method applied to voice signals where the higher fi-equencies are reallocated to a lower range where the human ear is most sensitive. Each fi-equency component of a voiced/unvoiced speech undergoes a certain translation to a lower frequency by an amount which depends on its position on the frequency axis. Higher frequencies would have a much higher shifting factor than the lower ones. Due to this frequency translation mechanism, the speech spectrum will have most of its power distributed among the lower frequency tones. Because of its predominance of lower frequency components, this system could be found usefi.d in hearing aids, voice mail applications, telephone, or radio communications. This speech processor is being implemented on the Analog Devices ADSP-2101 fixed-point digital signal processor, while the main principle is illustrated using MATLA13.
Even though the normal human hearing range can extend to about 15 KHz or more, speech is confined mainly in the range which extends fi-om 100 Hz to about 5 KHz1. In elderly people, the ear looses its sensitivity to the high-pitched tones, resulting in poor and inadequate hearing. Since, even with age, the ear is more sensitive to lower fi-equencies, it would be desirable to shift most of the energy contained in the upper frequencies to a lower range, provided the speech inilormation is fairly preserved. This is in contrast with conventional hearing aids, for example, which rely mainly on amplification and filtering. Excessive amplification of both low and high frequencies could result in severe damage of the remaining hearing of an elderly person. By forcing the speech power to be reallocated within the low and mid-frequencies, unnecessary amplification can be avoided.
This paper addresses the alteration of the frequency spectrum of speech signals using a nonlinear frequency mapping technique, where the lower fi-equencies undergo no change, or very little, in this transformation process, whereas the higher tones are subject to more shifiing. The general principle consists of computing the frequency spectrum of spoken words using the FFT (Fast Fourier Transform) algorithm, manipulating that spectrum so that most of the energy is transferred to a lower frequency range by application of the frequency mapping technique to be presented here, and finally taking the inverse FFT of the transformed spectrum.
This method is still being implemented on the Analog Devices ADSP-21O 1 fixed-point digital signal processor. The implementation is the most critical part since the system uses nonlinear frequency mapping, and must respond to red-time situations, where, for example, a spoken word into the system must be heard back altered, but understandable, reflecting the change in its frequency spectrum. The challenge remains in the choice
$&”-:1996 ASEE Annual Conference Proceedings } ‘..+,p+$ .
Bouktache, E., & Sekhar, C. R., & Farook, O. (1996, June), Alteration Of The Frequency Profile Of Voice Signals Paper presented at 1996 Annual Conference, Washington, District of Columbia. 10.18260/1-2--5884
ASEE holds the copyright on this document. It may be read by the public free of charge. Authors may archive their work on personal websites or in institutional repositories with the following citation: © 1996 American Society for Engineering Education. Other scholars may excerpt or quote from these materials with the same citation. When excerpting or quoting from Conference Proceedings, authors should, in addition to noting the ASEE copyright, list all the original authors and their institutions and name the host city of the conference. - Last updated April 1, 2015