Marios Fourakis

SPEECH RECOGNITION AT SIMULATED SOFT, CONVERSATIONAL, AND RAISED-TO-LOUD VOCAL EFFORTS BY ADULTS WITH COCHLEAR IMPLANTS

Ten postlinguistically deaf adults who used the Nucleus Cochlear Implant System and SPEAK speech coding strategy responded to vowels, consonants, words, and sentences presented sound-only at 70, 60, and 50 dB sound-pressure level. Highest group mean scores were at a raised-to-loud level of 70 dB for consonants (73%), words (44%), and sentences (87%); the highest score for vowels (70%) was at a conversational level of 60 dB. Lowest group mean scores were at a soft level of 50 dB for vowels (56%), consonants (47%), words (10%), and sentences (29%); all except subject 7 had some open-set speech recognition at this level. For the conversational level (60 dB), group mean scores for sentences and words were 72% and 29%, respectively. With this performance and sound-pressure level, it was observed that these subjects communicated successfully in a variety of listening situations. Given these subjects speech recognition scores at 60 dB and the fact that 70 dB does not simulate the vocal effort used in everyday speaking situations, it is suggested that cochlear implant candidates and implantees be evaluated with speech tests presented at 60 dB instead of the customary 70 dB sound-pressure level to simulate benefit provided by implants in everyday life. Analysis of individuals scores at the three levels for the four speech materials revealed different patterns of speech recognition among subjects (e.g., subjects 1 and 5). Future research on the relation between stimuli, sound processing, and subjects responses associated with these different patterns may provide guidelines to select parameter values with which to map incoming sound onto an individuals electrical dynamic range between threshold and maximum acceptable loudness level to improve speech recognition.

Identification of speech by cochlear implant recipients with the multipeak (MPEAK) and spectral peak (SPEAK) speech coding strategies II. Consonants.

OBJECTIVEnThe major objective of this study was to evaluate differences in consonant recognition with the Multipeak (MPEAK) and the Spectral Peak (SPEAK) speech coding strategies of the Nucleus-22 Cochlear Implant System. This objective was addressed by comparison of acoustic and electrode activation analyses of consonants with cochlear implant recipients responses to these same consonant tokens when they used the two speech coding strategies.nnnDESIGNnNine subjects identified 14 English consonants with the MPEAK and SPEAK speech coding strategies. These strategies were compared with an ABAB design. Evaluation occurred during two weekly sessions after subjects used each strategy for at least 3 wk in everyday life.nnnRESULTSnGroup medial consonant [aCa] identification scores with the SPEAK strategy were significantly higher than with the MPEAK strategy (76.2% versus 67.5%; p < 0.001). This improvement was largely due to the significant increase in information transmitted for the place feature (p < 0.001) through accurate tracking of second formant transitions and spectrally specific stimulation patterns to differentiate [s] from [symbol see text] and [n] from [m], and the stop consonant bursts. For this reason, more nasal consonants were correctly identified with SPEAK, but there also were more non-nasal error responses when the nasal murmur was of unusually low amplitude. Consequently, significantly less information was transmitted for the nasality feature with SPEAK than MPEAK (p < 0.001).nnnCONCLUSIONSnElectrical stimulation with the SPEAK strategy provided better spectral representation of the stop consonant bursts, tracking formant transitions into the following vowel, frication in the consonant [symbol see text], and the formants for the nasals [m] and [n] than with the MPEAK strategy. The marked improvement in recognition of the velar consonants, [g] and [k], which cannot be seen during speechreading, should allow greater ease and accuracy of communication with SPEAK than MPEAK.

Acoustic characteristics of Greek vowels.

Five male speakers produced the vowels of Greek at slow and fast tempo, in lexically stressed and unstressed syllables, and in lexically stressed syllables of words appearing in focus position. Duration, fundamental frequency (F₀), amplitude, and the frequencies of the first (F1) and second formant (F2) were measured. The effects on these variables of the phonemic category of the vowel, tempo, stress, and focus were examined. The results indicated that the vowel system of Greek follows universal tendencies in terms of duration but not in terms of F₀ and amplitude. Vowels in focus position, when plotted by their F1 and F2 frequencies, defined a vowel space larger than that defined by vowels in any other condition.

The acoustic vowel space of modern Greek and German.

The spectral characteristics of vowels in Modern Greek and German were examined. Four speakers of Modern Greek and three speakers of German produced four repetitions of words containing each vowel of their native language. Measurements of the fundamental frequency and the first three formants were made for each vowel token. These measurements were then transformed into log frequency ratios and plotted as points in the three-dimensional auditory-perceptual space proposed by Miller (1989). Each vowel token was thus represented by one point, and the points corresponding to each vowel category were enclosed in three-dimensional target zones. For the present corpus, these zones differentiate the five vowels of Modern Greek with 100% accuracy, and the fourteen vowels of German with 94% accuracy. Implications for the distribution of common vowels across languages as a function of vowel density are discussed.

On discrete changes in the acquisition of the alveolar/velar stop consonant contrast

Les As. proposent un commentaire sur larticle de Thomas Berg Sound change in child language : a study of inter-word variation paru en 1995 dans Language and Speech. Dans cet article, Berg conclut que lenfant quil etudie acquiert graduellement le contraste alveolaire/velaire en 1 e position et quil ne presente pas de contraste cache. Les As. montrent ici quune chose apparemment discrete comme la substitution alveolaire/velaire dans les consonnes occlusives peut en realite etre completement differente et presenter des contrastes caches, si lon ajoute aux jugements auditifs et perceptuels employes par Berg, des analyses acoustiques et/ou articulatoires

An Acoustic Study of the Effects of Tempo and Stress on Segmental Intervals in Modern Greek

Changes in stress and rate of speech have been shown to have different effects on the durations of speech intervals. We examined the effects of such changes on syllable and intrasyllabic segment durations in the word-initial syllable of three-syllable words in Modern Greek, and in conflict with previously reported results for other languages, found rate and stress to have equal effects. We propose that languages may differ in terms of temporal programming in the same sense as they differ in terms of syntactic or phonological rule systems.

Bi-syllabic, Modern Greek word lists for use in word recognition tests

The development of a word recognition test for Modern Greek, which is comprised of three fifty-word lists, is described herein. The development was guided by four principles: 1) use of the shortest words possible (two syllables for Greek) 2) use of highly frequent words 3) phonetic balance and 4) appropriate balance of first and second syllable stress. The lists were recorded by one male and one female native speakers. Thirty-seven native speakers of Greek listened to all words by both speakers. Across lists, the mean correct identification score was 97.9% for the female voice (95% confidence interval 96.97 to 98.84) and 96.5% (95% confidence interval 95.31 to 97.77) for the male voice. This small difference was statistically significant (p<.01) and concentrated on words with first syllable stress. In future work, these recordings can be used in adult tests of speech perception and can be modified for tests of central auditory processing. Sumario Se describe el desarrollo de una prueba de discriminación en griego moderno a partir de tres listas de cincuenta palabras. Este desarrollo se basó en tres principios: 1. Uso de las palabras más cortas posibles en Griego (bisilábicos); 2. Uso de palabras altamente frecuentes; 3. Balance fonético 4. Balance apropiado entre el acento inicial y final. Las palabras fueron grabadas por un hombre y una mujer nativos del idioma. 37 personas nativos del idioma griego escucharon todas las palabras con ambas voces. En todas las listas, el porcentaje de identificación correcto fue del 97% para la voz femenina (Intervalo de confianza de 95% de 96.97 a 98.84) y 96.5% (intervalo de confianza del 95%, de 95.31 a 97.7%) para la voz masculina. Esta pequeña diferencia fue estadísticamente significativa (p<.01) y se concentró en las palabras con acento en la primera sílaba. Estas grabaciones pueden ser utilizadas en futuros trabajos para pruebas de percepción del lenguaje en adultos y pueden ser modificadas para pruebas de procesamiento central auditivo.

Effects of Metrical Foot Structure on Syllable Timing

The durations of syllabic intervals in sentences with different rhythmic structure were examined. Rhythmic structure was defined as the organization of stressed and unstressed syllables into metrical feet — in this case, iambs and anapests. Each sentence contained three metrical feet, and each foot could be either an iamb or an anapest; hence, there were eight sentences of different rhythm types. Six. native speakers of American English each read 24 versions of each of the 8 rhythm types (192 sentences). The utterances were recorded and the durations of each syllable of each foot were measured from sound spectrograms. Statistical analysis indicated that the durations of some syllables were affected by the structure of the foot that contained them. There was also, in most cases, a significant shortening of the stressed syllable of a foot when the following foot was an anapest rather than an iamb. The results were interpreted as indicating that the process by which speech is produced is not strictly concatenative but involves planning which extends to at least two metrical feet at a time.

An acoustic metric for assessing change in vowel production by profoundly hearing‐impaired children

The purpose of this study was to investigate the feasibility of developing an acoustic metric to assess vowel production in profoundly hearing-impaired children. The approach taken was to develop a metric from acoustic analysis of vowel productions and then compare it with the perceptual ratings of the same productions by listeners. Speech samples were collected from three profoundly hearing-impaired children participating in a longitudinal study that investigated the effectiveness of assistive listening devices upon speech development. The metric used the extracted fundamental and first, second, and third formant frequencies to represent the tokens as points in a three-dimensional auditory-perceptual space modeled after earlier work by Miller [J. Acoust. Soc. Am. 85, 2114-2134 (1989)]. Euclidean distances were determined between each point and the intended vowel, which was represented by coordinates taken from the Peterson and Barney [J. Acoust. Soc. Am. 24, 175-184 (1952)] data for children. The data suggest that the three-dimensional metric provides significant correlations between production and perception.

Why [spa] not [psa]? On the perceptual salience of initial /s/‐stop and stop‐/s/ sequences.

Initial /s/‐stop clusters occur frequently in the world’s languages, but initial stop‐/s/ clusters are relatively infrequent. Furthermore, there appear to be no languages that contain initial stop‐/s/ clusters, but not /s/‐stop clusters, while the reverse is not true [Morelli, (1999) and (2003)]. This study aims at uncovering a perceptual explanation for these patterns by examining the salience of initial /s/‐stop and stop‐/s/ clusters in Greek, where both sequences are common. Twenty naive Greek adult listeners identified syllables beginning with /sp/, /st/, /sk/, /ps/, /ts/, or /ks/, in two vowel contexts, /a/ and /i/, in real words spoken by ten Greek adult native speakers. The syllables were mixed with parts of Greek multitalker babble using SNRs of −6, 0, and +6 dB and presented to listeners for identification. Results showed significantly poorer identification for the /ps/ and /ks/ clusters than the /ts/ and /s/‐stop clusters, particularly in the −6 and 0 SNRS. There was also a significant interacti...