Sound Spectrum

4 Key excerpts on "Sound Spectrum"

• 

  eBook - ePub

  Phonetics

  The Science of Speech

  • Martin J Ball, Joan Rahilly(Authors)
  • 2014(Publication Date)
  • Routledge

    (Publisher)

  In phonetics research, the notion of resonance allows us to classify speech sounds using spectra. Spectra are defined by Fry (1979, p. 58) as ‘statement[s] of what frequencies are to be found in the mixture and what their relative amplitudes are’. Later, we will examine the components of spectra in detail but, for the moment, it will suffice to state that spectral measurement allows us to measure the fundamental frequency, along with the harmonics that are present in speech.

  Basic methods in acoustic analysis—————

  As we stated at the beginning of this chapter, acoustic phonetics aims to provide a quantifiable record of speech events using instrumental techniques. These techniques are usually based around computer software packages, often interfaced with hardware which produce visual representations of and statistical measurements of speech. Although this explanation seems to represent acoustic phonetics as an exact science, we must point out that this is not the case. For instance, if we were using tape-recorded material as input, it may be the case that the quality of the recording is less than ideal (because of extraneous background noise, or inappropriate recording levels, for instance). It may also be that the chosen acoustic program fails to analyse the data in the best way, either because of its own design faults, or because we ourselves have failed to specify the correct settings for data sampling.

  Because of the potential difficulties outlined above, it is important to familiarize ourselves with good practice in recording techniques, and we should also equip ourselves with a solid understanding of the characteristics and methods of whatever acoustic analysis program we happen to be using. In addition to these prerequisites to acoustic analysis, we must also acquire a firm grasp of basic procedures in acoustics, and of how acoustic representations relate to articulatory characteristics of speech. The rest of this chapter concentrates on these aspects.

  Sign up to read

  Learn more about book
• 

  eBook - PDF

  Introduction to English Phonetics and Phonology

  • Ulrike Gut(Author)
  • 2014(Publication Date)
  • Peter Lang GmbH, Internationaler Verlag der Wissenschaften

    (Publisher)

  You can see that only some of the frequencies have a high intensity (in particular the fundamental frequency at 185 Hz, which has the highest intensity). This is caused by the particular position of the speaker's tongue during the articulation of this vowel - a high tongue position with the front of the tongue elosest to the roof of the mouth. Another way of visualizing the intensity of the different frequencies of voiced speech sounds is the spectrogram. Speech analysis software usually creates spectrograms from recordings by performing a Fourier analysis (named after Jean Baptiste Fourier, who showed in the early 19 th century that complex waves can be analysed as the sum of their sine wave components). With this analysis, the individual frequencies of each speech sound and their corresponding intensities can be made visible. Another technique for examining the spectral properties of sounds, which is offered by many speech analysis packages, is the linear prediction coefficient analysis (LPC). Acoustic properties of English 147 40 N "' 20 '"

  Sign up to read

  Learn more about book
• 

  eBook - PDF

  Contemporary Issues in Experimental Phonetics

  • Norman Lass(Author)
  • 2012(Publication Date)
  • Academic Press

    (Publisher)

  Spectrographic analysis (Joos, 1948; Potter, Kopp, and Green, 1947) and speech synthesis (Liberman, 1957) showed that patterns of speech important to its perception lay not in its waveform but in its time-varying spectrum as revealed by the spectrogram. We may imagine, therefore, an early stage of the auditory display, soon after cochlear analysis, as the neural correlate of a spectrogram. Notice in Figure 8.1: regions of high energy concentration (formants, usually labeled from bottom to top as Fl, F2, F3); different formant patterns associated with the vowels of read and book, for example; intervals of silence during stop consonant closure; a sharp scatter of energy (noise burst) upon release of the voiceless stop in to, and fainter bursts following release of the voiced stops in began; rapid formant movements (transitions) as articulators move into and out of vowels; a nasal formant (between Fl and F2) at the end of began; a broad band of noise associated with the fricative of she; and finally, regular vertical striations, reflecting a series of glottal pulses, from which fundamental frequency can be derived. A later, perhaps cortical, stage of auditory analysis may entail detection of just such features in the spectrographic display. Whether there are acoustic feature analyzers specially tuned to speech is an open question that we consider later. In any case, the signal has not been transformed yet into the message, and, indeed, may have passed through the same processes as any other auditory input. s h e b e g a n t o r e a d h e r b o o k Figure 8.1 Spectrogram of a natural utterance: She began to read her book. Frequency is plotted against time, with relative intensity represented by degree of blackness. 246 Michael Studdert-Kennedy The phonetic level is abstract in the sense that its output is a set of properties not inherent in the signal.

  Sign up to read

  Learn more about book
• 

  eBook - ePub

  A Field Manual for Acoustic Phonetics

  • Joan L. G. Baart(Author)
  • 2018(Publication Date)
  • SIL International

    (Publisher)

  section 3.2.5 ). On the other hand, we can take a good look at the timing of events. The boundaries between the two vowels and the consonant can be seen (the formants belong to the vowels; the middle section without formants belongs to the plosive). Furthermore, we are able to discern the individual voicing pulses; they are seen as vertical dark lines that occur one after another during the vowels. No pulses are seen during the [p], which is voiceless.

  For most practical purposes, a speech researcher will inspect broad-band rather than narrow-band spectrograms. One reason is that the interest of the researcher is often in the formants, rather than in the separate harmonics. The detection of formants is actually facilitated by the blurring of information in the dimension of frequency. Another reason is that one often wants to see with some accuracy how acoustic parameters change over time, and for this purpose a good resolution in the time domain is required.

  3.1.4 Spectra

  A spectrum is a two-dimensional representation, with frequency displayed along the horizontal axis, and intensity along the vertical axis. A spectrum shows for a section of a sound wave what frequency components are present and what the relative intensities (strengths) of these components are. While a spectrogram gives a dynamic picture of a stretch of sound (it shows how acoustic parameters change over time), a spectrum gives a static picture: it has no time dimension.

  As with spectrograms, the appearance of a spectrum depends to an important extent on the size of the time window that is chosen for analysis. If a wave is periodic and the analysis window is smaller than the size of one cycle of the wave, then the analysis will be based on incomplete information. For the sake of precision, one would like to calculate a spectrum over as large a stretch of sound as possible. However, if one is interested in the spectral properties of only one vowel or consonant, one is constrained by the duration of that vowel or consonant. In fact, one will often be confined to a section in the middle of a segment, where the sound is maximally stationary, as the peripheral regions of speech sounds are often heavily influenced by neighboring sounds.

  Sign up to read

  Learn more about book