![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
(For future reference, I am now trying to keep these shorter than previously, as that is the point of a reading journal. If I want all of the information, I can go read the original.)
Remez, Robert E., and Rubin, Philip E. (1990). On the perception of speech from time-varying acoustic information: Contributions of amplitude variation. Perception and Psychophysics: 48(4), 313-325.
Remez & Rubin are at it again with their wacky sine-wave antics. In earlier studies, they'd claimed that patterns of frequency changes were the only aspect of speech preserved in their synthesis, and thus these frequency variations over time must be the only relevant thing when it comes to transcribing speech. However, they point out, this isn't the case: amplitude information is preserved, too. To rectify this, they did three little experiments to investigate the relative importance of frequency and amplitude information in sine-wave speech. In three experiments, they show that "fine-grain" changes in the amplitude of these speechlike sounds don't make any difference in subjects' accuracy in transcribing the sentences, but that obstruction of larger changes - for instance, the obliteration of natural silences surrounding consonants like "t" or "d" - causes serious deterioration of accuracy. Surprisingly, they also show that syllable counting in a sentence is based more on frequency shifts than amplitude changes; when amplitude is kept constant or varied inappropriately on top of a natural-frequency sentence, syllable counting is unimpaired, but when the inverse is true, more errors are made.
These boys get mad props for challenging one of their own conclusions. I am suitably impressed.
--
This article has interested me in the process of amplitude modulation. I played around with Sound Forge synthesis, and noticed that amplitude-modulating a simple sinewave tone obliterates the carrier frequency from the signal entirely - you can't hear it, it doesn't show up in FFT analysis, it's just gone. This strikes me as quite weird. What you do hear isn't even the modulating frequency - it's some bastard lovechild of the two which seems to be a chord made up of carrier + modulator and carrier - modulator. Weeeeeeird...
Remez, Robert E., and Rubin, Philip E. (1990). On the perception of speech from time-varying acoustic information: Contributions of amplitude variation. Perception and Psychophysics: 48(4), 313-325.
Remez & Rubin are at it again with their wacky sine-wave antics. In earlier studies, they'd claimed that patterns of frequency changes were the only aspect of speech preserved in their synthesis, and thus these frequency variations over time must be the only relevant thing when it comes to transcribing speech. However, they point out, this isn't the case: amplitude information is preserved, too. To rectify this, they did three little experiments to investigate the relative importance of frequency and amplitude information in sine-wave speech. In three experiments, they show that "fine-grain" changes in the amplitude of these speechlike sounds don't make any difference in subjects' accuracy in transcribing the sentences, but that obstruction of larger changes - for instance, the obliteration of natural silences surrounding consonants like "t" or "d" - causes serious deterioration of accuracy. Surprisingly, they also show that syllable counting in a sentence is based more on frequency shifts than amplitude changes; when amplitude is kept constant or varied inappropriately on top of a natural-frequency sentence, syllable counting is unimpaired, but when the inverse is true, more errors are made.
These boys get mad props for challenging one of their own conclusions. I am suitably impressed.
--
This article has interested me in the process of amplitude modulation. I played around with Sound Forge synthesis, and noticed that amplitude-modulating a simple sinewave tone obliterates the carrier frequency from the signal entirely - you can't hear it, it doesn't show up in FFT analysis, it's just gone. This strikes me as quite weird. What you do hear isn't even the modulating frequency - it's some bastard lovechild of the two which seems to be a chord made up of carrier + modulator and carrier - modulator. Weeeeeeird...