Peter W. Lenz

Intelligibility of bandpass filtered speech: Steepness of slopes required to eliminate transition band contributions

Despite the recognition that the steepness of filter slopes can play an important role in the intelligibility of bandpass speech, there has been no systematic examination of its importance. The present study used high orders of finite impulse response (FIR) filtering to produce slopes ranging from 150 to 10,000 dB/octave. The slopes flanked 1/3-octave passbands of everyday sentences having a center frequency of 1500 Hz (the region of highest intelligibility for the male speakers voice). Presentation levels were approximately 75 and 45 dB. No significant differences were found for the two presentation levels. Average intelligibility scores ranged from 77% at 150 dB/octave down to the asymptotic intelligibility score of 12% at 4800 dB/octave. These results indicate that slopes of several thousand dB/octave may be required for accurate and unambiguous specification of the range of frequencies contributing to intelligibility of filtered speech. In addition, the extremely steep slopes are needed to ensure that none of the spectral components contributing to intelligibility has its relative importance diminished by spectral tilt.

Intelligibilities of 1-octave rectangular bands spanning the speech spectrum when heard separately and paired.

There is a need, both for speech theory and for many practical applications, to know the intelligibilities of individual passbands that span the speech spectrum when they are heard singly and in combination. While indirect procedures have been employed for estimating passband intelligibilities (e.g., the Speech Intelligibility Index), direct measurements have been blocked by the confounding contributions from transition band slopes that accompany filtering. A recent study has reported that slopes of several thousand dBA/octave produced by high-order finite impulse response filtering were required to produce the effectively rectangular bands necessary to eliminate appreciable contributions from transition bands [Warren et al., J. Acoust. Soc. Am. 115, 1292-1295 (2004)]. Using such essentially vertical slopes, the present study employed sentences, and reports the intelligibilities of their six 1-octave contiguous passbands having center frequencies from 0.25 to 8 kHz when heard alone, and for each of their 15 possible pairings.

Enhancing intelligibility of narrowband speech with out-of-band noise: Evidence for lateral suppression at high-normal intensity

Previous studies have shown that the intelligibility of filtered speech can be enhanced by filling stopbands with noise. The present study found that this enhancement occurred only when speech intensity was sufficiently high to degrade performance. Intelligibility decreased by about 15% when narrowband speech was increased from 45 to 65 dBA (corresponding to broadband speech levels of about 60 and 80 dBA), and decreased by 20% at a level of 75 dBA. However, when flanking bands of low-pass and high-pass filtered white noise were added at spectrum levels of -40 to -20 dB relative to the speech, intelligibility of the 75-dBA speech band increased by about 13%. Additional findings confirm that this enhancement of intelligibility depends upon out-of-band stimulation, in agreement with theories proposing that lateral suppressive interactions extend the dynamic range of intensity coding by counteracting effects of auditory-nerve firing-rate saturation at high signal levels.

Relative contributions of passband and filter skirts to the intelligibility of bandpass speech: Some effects of context and amplitude

Warren et al. (1995) reported over 90% intelligibility for everyday sentences reduced to a 1/3-octave band (center frequency 1,500 Hz, slopes 100 dB/octave, slow-rms peak levels 75 dB). To investigate the basis of this high intelligibility, Warren and Bashford (1999) partitioned the sentences. Surprisingly, the rectangular 1/3-octave passband had only 24% intelligibility, whereas the filter skirts separated by a 1/3-octave notch had an intelligibility of 83%, despite their severe spectral tilts. Experiment 1 of the present study substituted monosyllabic words for sentences. Wholeband intelligibility was 26%, the passband 4%, and the filter skirts 16%. Experiment 2 measured intelligibility for 1/3-octave sentences having peak levels ranging from 85 down to 35 dB. Whole band intelligibility ranged from 90% to 68%, and the filter skirt pairs had from two to four times the passbands intelligibility (which did not vary significantly with level). Hence, steep (100 dB/octave) filter skirts make the dominant contribution to intelligibility of nominally 1/3-octave speech across a wide range of presentation levels.

Polling the effective neighborhoods of spoken words with the verbal transformation effect

Studies of the effects of lexical neighbors upon the recognition of spoken words have generally assumed that the most salient competitors differ by a single phoneme. The present study employs a procedure that induces the listeners to perceive and call out the salient competitors. By presenting a recording of a monosyllable repeated over and over, perceptual adaptation is produced, and perception of the stimulus is replaced by perception of a competitor. Reports from groups of subjects were obtained for monosyllables that vary in their frequency-weighted neighborhood density. The findings are compared with predictions based upon the neighborhood activation model.

An alternative to the computational Speech Intelligibility Index estimates: direct measurement of rectangular passband intelligibilities.

The need for determining the relative intelligibility of passbands spanning the speech spectrum has been addressed by publications of the American National Standards Institute (ANSI). When the Articulation Index (AI) standard (ANSI, S3.5, 1969, R1986) was developed, available filters confounded passband and slope contributions. The AI procedure and its updated successor, the Speech Intelligibility Index (SII) standard (ANSI, S3.5, 1997, R2007), cancel slope contributions by using intelligibility scores for partially masked highpass and lowpass speech to calculate passband importance values; these values can be converted to passband intelligibility predictions using transfer functions. However, by using very high-order digital filtering, it is now possible to eliminate contributions from filter skirts and produce rectangular passbands. Employing the same commercial recording and the same one-octave passbands published in the SII standard (Table B.3), the present study compares Rectangular Passband Intelligibility (RPI) with SII estimates of intelligibility. The directly measured RPI differs from the computational SII predictions. Advantages resulting from direct measurement are discussed.

Evoking biphone neighborhoods with verbal transformations: illusory changes demonstrate both lexical competition and inhibition.

When a recorded verbal stimulus repeats over and over, perceptual changes occur and listeners hear competing forms. These verbal transformations (VTs) were obtained for a phonemically related set of 24 consonant-vowel syllables that varied widely in frequency-weighted neighborhood density (FWND). Listeners initial transformations involving substitution of consonants versus vowels were strongly correlated with the lexical substitution neighborhood [r=+0.82, p<0.0001]. Interestingly, as stimulus FWND increased, average time spent hearing illusory forms substantially decreased [r=-0.75, p<0.0001]. These results suggest that VTs not only reveal underlying competitors, but also provide a highly sensitive measure of lexical inhibition.

Intelligibility of dual rectangular speech bands: implications of observations concerning amplitude mismatch and asynchrony

The present study examines the integration of information present in different spectral regions of speech using two 1/3-octave bands of everyday sentences (center frequencies 1- and 3-kHz). Nearly vertical slopes were employed (4000- order finite impulse response filtering) to avoid the major contribution to intelligibility made even by conventionally steep slopes (e.g., 100 dB/octave) (see [J. Acoust. Soc. Amer. 108 (2000) 1264]). Heard alone at 75 dB, the rectangular band intelligibilities were 5% (1 kHz) and 10% (3 kHz); heard together, their score was 77%. Conformity to the normal spectral profile was not required for this remarkably high degree of synergy: When the 3-kHz band was kept at 75 dB and the 1-kHz bands level was decreased systematically, intelligibility remained unchanged from 75 to 45 dB (intensity ratio of 1000:1). But when the rectangular bands were kept at their normal levels, and one band was delayed relative to the other, intelligibility dropped to half with a misalignment of only about 35 ms (approximately half the duration of the average phoneme); scores dropped further, approaching that of a single band when asynchrony approximated average phonemic durations.

Maintaining intelligibility at high speech intensities: Evidence of lateral inhibition in the lower auditory pathway

Three experiments examined the intelligibility enhancement produced when noise bands flank high intensity narrowband speech. Enhancement was unaffected by noise gating (experiment 1), ruling out peripheral adaptation as a source, and was also unaffected by interaural decorrelation of noise bands flanking diotic speech (experiment 2), indicating that enhancement occurs prior to binaural processing. These results support previous suggestions that intelligibility loss at high intensities is reduced by lateral inhibition in the cochlear nuclei. Results from a final experiment suggest that this effect is only ipsilateral, implicating a specific population of inhibitory neurons.

Critical bandwidth speech: Arrays of subcritical band speech maintain near-ceiling intelligibility at high amplitudes

Removal of transition bands from narrow speech passbands through very steep filtering has made it possible to isolate and determine (for the first time) intelligibility of critical bandwidth as well as subcritical bandwidth speech. These rectangular bands have unique intelligibilities when heard singly, paired, or in various multiband arrays spanning the speech spectrum. Thus, a particular sparse spectral array of unfamiliar everyday sentences in this study has demonstrated 98% intelligibility up to 100 dB. Some theoretical and practical applications are suggested.