1, 6.551J/HST714J Lecture A' ACOUSTICS OF SPEECH AND HEARING /ay V/ * 2 Features for consonants in English p t k b d g cont strid. lips blade body ant nas voic + + + + + + f + V S z + + + + + + + I s z + + + + + +I + -/+ 0 6 m n ju -'+ + ± + + I- + + + + + + + + + + + + + + + + _ + + + _ + + + Table 1. Articulator-free features for some consonants in English t,d continuant s,z 0,6 + + + sonorant strident + 1 n + w Y J. Acoust. Sc. An Stevens, K.N.:Distinctive Features 67 Table 2. Seven articulators and the features that specify phonetic distinctions that can be made with each articulator. The articulators are divided into three groups: those that can form a constriction in the oral cavity, those that control the shape of the airway in the laryngeal and pharyngeal regions, and the aspect of vocal-fold adjustment relating to stiffness. lips [round] tongue blade [anterior] [distributed] [lateral] [rhotic] tongue body [high] [low] [back] - soft palate [nasal] pharynx [advanced tongue root] glottis [spread glottis] [constricted glottis] vocal folds [stiff vocal folds] Th voicing: +: bdgv6zij -: ptkfOsc coronal (blade) +: tdOszszcjn -: pkbgfvmj strident +: szizjfv -: tdgjO 6mnpkb English plural az kiss bush fez rouge church judge s hat cake cap cuff z cab cad cog cave quay eye ham hen Nasal Assimilation land lamp laik Suffix /jan/ commune hell communion confess + jan deride contrite Egypt divide confession derision contrition Egyptian division hellion million J. Acoust. Sc. Arn Stevens, K.N.:Distinctive Features 73 Table 5. Lexical representations for the words help, debate and wagon. The syllable structure of each word is schematized at the top ((5 = syllable, o = onset, r = rime). r a r oar r),O/ r 0 d a vowel b e + t ae w + g + glide n o + + + stressed - reducible + p + + + + + + + 1 + + consonant h + - + - continuant sonorant - - + + strident lips + tongue blade + + + tongue body + + + + + round + anterior + + lateral + high + - low - - back - - + + + adv. tongue root + - + + - - - + spread glottis + nasal stiff vocal folds + - - + - + A 4 ^% n^^ Ott A^f c C LSPECTO: 256-pt DFT, smart AGC (STEVENS_DATA.COURSE.SPRINGOO] · I'I.~I . oAt nA '7An o^rs 6.4-ms Hamming window every 1 ms APR 11 2000 _. 4f afn Ann THEDEBATE .I . i ,l~~~~L hRj . I' I 3 a .Ii 1000 'r- -Aq~h 0 1- w .I I N "I 2 I . , =J i i iJ A - s .I 41m~Il F, 1~~~~~- - I ' 0 100 A 4A^ 200 300 400 500 600 TIME (ms) 700 800 900 7n OAA Ann 80 n 40 -< _ nnn "An A N nA a aAnn, j~~~ 6 aclcessc 7 chtb S ee_ uibi\it -Cov /xIouLL s ve eh Aan ts RATIO IN OCaELS -K,.tLO-NOISE Figure 2: Dependence of word identification score on speech to noise ratio for keywords spoken in sentences (five per sentence) and for the same words spoken in isolation (from Miller, Heise, and Lichten, 1951). SIGNAL-TO-NOtSE RATIO TI&L la2taWitr o =o0o0ylabk a a& IN DEGBLI d AlJe Wat-g ob t soawdwr a) as Figure 1: Dependence of intelligibility on the number of items (monosyllabic words) in the test vocabulary (from Miller, Heise, and Lichten, 1951). S n \ 1 P,5: IU -e A Ce, o IV iC,0 A eA-~ · I I ~-1' 1 P~p1 F- 75 U w PH/ I I I 0 50 I I I - Z W I U Ia w 25 I I I II * 0 I- 0 w U a: W S/N RATIO (dB) FIG. 4. PH and PL scores versus S/N ratio for the original 10 test forms (250 items of each type), for normally hearing young and old listeners. .&Mp les: PF: -The boatcXct PFL: WewerFe tk accs -&bbq boL ut th1i v CoSwloY matibCts FDOnr.YI!Ot)Yl ti in slabkes TABLE V. Confusion matrix for S/N- +6 db and frequency response of 200-6500 cps. p t k p I 162 8 38 10 270 6 55 14 171 5 5 1 1 I 2 2 f * k 5S f b t d f V s 3 5 3 m . 1 1 207 71 1 S 57 142 3 7. 232 1 1 b d t 3 2 239 2 214 206 64 14 194 3 3 3 2 4 10 4 m n 1 11 I '5 1 2 1 14 2 I 2 31 12 9 4 205 55 2 39 179 20 1 2 2 1 5 22 198 2 2 3 215 I 217 2 3 285 2 2 3 TABLE II. Confusion matrix for S/N- -6 db and frequency response of 200-6500 cps. p p A f e 5 f 80 71 66 43 64 84 55 76 107 18 19 8 1 12 17 S 6 1 b d 9 16 4 3 17 5 12 rs S 6 3 9 2 8 4 I 11 32 107 29 1 7 45 195 7 5 4 2 4 2 3 1 5 3 2 6 1 8 136 5 3 10 80 63 9 45 66 47 11 3 48 31 7 1 5 6 20 26 5 17 27 18 14 9 8 175 48 104 64 23 39 4 .6 $ 4 4 2 g 2 16 a 3 2 6 I I 1 b 1 M a d 1 4 I 4 g 3 tS 1 2 2 4 1 5 3 2 16 20 19 6 20 37 2 26 56 145 86 16 3 45 58 28 8 12 21 94 45 5 44 129 4 2 7 3 1 2 5 1 I 6 1 1 5 1 4 3 4 6 4 2 177 47 46 163 TALZE II. Confusion matrix for S/IN- -12 db and frequency response 200-6500 cp. p d g 65 74 62 22 20 22 19 24 18 6 22 16 11 14 .11 2 2 4 3 1 2 1 1 3 1 1 3 2 2 I 1 S 1 31 26 16 23 22 22 15 32 28 25 16 20 85 63 33 14 34 45 24 27 15 27 53 25 II 12 48 I1S 3 6 3 I 5 9 5 4 3 6 5 8 11 3 3 8 9 1 3 3 6 6 2 2 3 4 3 3 2 1 2 1 1 18 4 1 7 7 4 7 4 5 1 11 7 60 18 20 18 48 38 18 35 29 44 16 16 25 24 29 14 26 29 I 1 1 2 1 4 2 2 12 17 2 I 5 2 6 4. 3 14 6 5 2 8 7 37 53 23 7 20 31 29 30 23 25 27 23 71 50 24 9 16 33 19 7 I I 1 11 2 3 2 6 6 8 7 I1 1 I I # 5 b 3 57 42 6 3 k 's S5 51 64 50 I I m 8 5 4 3 7 S 1 2 4 2 1 2 6 14 38 20 9 10 10 12 9 14 23 40 39 4 S 26 77 14 13 3 5 9 6 6 14 1 1 9 109 84 60 145 immpile M a /0 ErSpeFigure2pol Figure 2 Discrete Output tinuous Gestures Sound Output /1 Figure 4 Universal component . . Phonological Representations Mo ! 16! Language specific component N1 PI - I discrete 4- - - A - - - - - -& 4 - el)!C, eI li LU ii continuous epresentation Overlap Vocal Tract m Sound output