A role for amplitude modulation phase relationships in speech rhythm perception

Leong, Victoria; Stone, Michael; Turner, Richard E.; Goswami, Usha

doi:10.1121/1.4883366

Cited by 55 publications

(120 citation statements)

References 59 publications

Supporting

Mentioning

107

Contrasting

Unclassified

Order By: Relevance

“…We also provide a Praat script for this purpose, which researchers interested in cochlear implant design, speech chimera, or amplitude envelope-based approaches to speech rhythm [20,21,22,23,24,25] may find useful. Depending on the need, this script can be easily modified into a procedure to be implemented in a bigger program, or saved as a button in the Praat object window.…”

Section: Resultsmentioning

confidence: 99%

A Praat-Based Algorithm to Extract the Amplitude Envelope and Temporal Fine Structure Using the Hilbert Transform

Dellwo

2016

Interspeech 2016

View full text Add to dashboard Cite

A speech signal can be viewed as a high frequency carrier signal containing the temporal fine structure (TFS) that is modulated by a low frequency envelope (ENV). A widely used method to decompose a speech signal into the TFS and ENV is the Hilbert transform. Although this method has been available for about one century and is widely applied in various kinds of speech processing tasks (e.g. speech chimeras), there are only very few speech processing packages that contain readily available functions for the Hilbert transform, and there is very little textbook type literature tailored for speech scientists to explain the processes behind the transform. With this paper we provide the code for carrying out the Hilbert operation to obtain the TFS and ENV in the widely used speech processing software Praat, and explain the basics of the procedure. To verify our code, we compare the Hilbert transform in Praat with a widely applied function for the same purpose in MATLAB ("hilbert(...)"). We can confirm that both methods arrive at identical outputs. AbstractA speech signal can be viewed as a high frequency carrier signal containing the temporal fine structure (TFS) that is modulated by a low frequency envelope (ENV). A widely used method to decompose a speech signal into the TFS and ENV is the Hilbert transform. Although this method has been available for about one century and is widely applied in various kinds of speech processing tasks (e.g. speech chimeras), there are only very few speech processing packages that contain readily available functions for the Hilbert transform, and there is very little textbook type literature tailored for speech scientists to explain the processes behind the transform. With this paper we provide the code for carrying out the Hilbert operation to obtain the TFS and ENV in the widely used speech processing software Praat, and explain the basics of the procedure. To verify our code, we compare the Hilbert transform in Praat with a widely applied function for the same purpose in MATLAB ("hilbert(...)"). We can confirm that both methods arrive at identical outputs.

show abstract

Section: Resultsmentioning

confidence: 99%

A Praat-Based Algorithm to Extract the Amplitude Envelope and Temporal Fine Structure Using the Hilbert Transform

Dellwo

2016

Interspeech 2016

View full text Add to dashboard Cite

show abstract

“…The reduction of temporal amplitude cues through normalization of vowel level may have reduced cues about the relative prominence of the sentence. However, this processing would have maintained the syllabic rhythm of the sentence, while other levels of the rhythmic hierarchy (Liberman & Prince, 1977) associated with slower stress patterning or faster phonetic cues would have been distorted (see Leong, Stone, Turner, & Goswami, 2014). It may be that restoring temporal amplitude cues did not improve the intelligibility of prominent words in this study due to the preservation of syllabic rate across signal processing conditions.…”

Section: Discussionmentioning

confidence: 99%

Indexical properties influence time-varying amplitude and fundamental frequency contributions of vowels to sentence intelligibility

Fogerty

2015

Journal of Phonetics

View full text Add to dashboard Cite

The present study investigated how non-linguistic, indexical information about talker identity interacts with contributions to sentence intelligibility by the time-varying amplitude (temporal envelope) and fundamental frequency (F0). Young normal-hearing adults listened to sentences that preserved the original consonants but replaced the vowels with a single vowel production. This replacement vowel selectively preserved amplitude or F0 cues of the original vowel, but replaced cues to phonetic identity. Original vowel duration was always preserved. Three experiments investigated indexical contributions by replacing vowels with productions from the same or different talker, or by acoustically morphing the original vowel. These stimulus conditions investigated how vowel suprasegmental and indexical properties interact and contribute to intelligibility independently from phonetic information. Results demonstrated that indexical properties influence the relative contribution of suprasegmental properties to sentence intelligibility. F0 variations are particularly important in the presence of conflicting indexical information. Temporal envelope modulations significantly improve sentence intelligibility, but are enhanced when either indexical or F0 cues are available. These findings suggest that F0 and other indexical cues may facilitate perceptually grouping suprasegmental properties of vowels with the remainder of the sentence. Temporal envelope modulations of vowels may contribute to intelligibility once they are successfully integrated with the preserved signal.

show abstract

“…Both sensations show a bandpass characteristic with peaks at 4 Hz for FS and 70 Hz for roughness. The range of modulations below 20 Hz has been shown to be of special interest for speech intelligibility [6,7] as well as for the perception of rhythm, which is related to the average syllable rate at AMs of around 4 Hz [8].…”

Section: Introductionmentioning

confidence: 99%