Speech Communications Chapter 7 Speech Communications The Nature

  • Slides: 15
Download presentation
Speech Communications Chapter 7

Speech Communications Chapter 7

Speech Communications The Nature of Speech Criteria for Evaluating Speech Components of Speech Communication

Speech Communications The Nature of Speech Criteria for Evaluating Speech Components of Speech Communication System Synthesized Speech

The Nature of Speech 1/2 發聲: 呼吸系統, Articulators Types of Speech Sound ¡Phoneme (音素)

The Nature of Speech 1/2 發聲: 呼吸系統, Articulators Types of Speech Sound ¡Phoneme (音素) − shortest segment of speech if change → meaning change ¡分類: 母音 (vowel), 子音 (consonant) 雙母音 (diphthongs) ¡Phoneme →Syllable →Word → Sentence

The Nature of Speech 2/2 Depicting Speech ¡Waveform, Spectrum ¡Sound spectrogram Fig 8 -1

The Nature of Speech 2/2 Depicting Speech ¡Waveform, Spectrum ¡Sound spectrogram Fig 8 -1 Intensity of Speech ¡Average intensity (speech power): 母音>子音 ¡Intelligibility: 子音較重要 Frequency Composition of Speech ¡低頻: 男>女 Fig 8 -2 ¡Shouting: frequency上升

Criteria for Evaluating Speech Intelligibility (能解度) ¡方法 − Repeat 呈現的聲音 − 回答問題 ¡Test −

Criteria for Evaluating Speech Intelligibility (能解度) ¡方法 − Repeat 呈現的聲音 − 回答問題 ¡Test − Nonsense syllables − Isolated words (phonetically balanced, PB) − Sentences Speech quality (Naturalness) ¡Preference

Components of Speech Communication System Speaker Message Transmission System Noise Hearer

Components of Speech Communication System Speaker Message Transmission System Noise Hearer

Components of Speech Communication System Speaker ¡Enunciation (清晰的聲音) ¡Superior Speakers − Longer syllable duration

Components of Speech Communication System Speaker ¡Enunciation (清晰的聲音) ¡Superior Speakers − Longer syllable duration − Greater intensity − More total time with speech sounds − Frequencies varied 1/7

Components of Speech Communication System Message ¡Phoneme Confusion − DVPBGCET, FXSH, KJA, MN −

Components of Speech Communication System Message ¡Phoneme Confusion − DVPBGCET, FXSH, KJA, MN − Avoid single letters, Word-spelling alphabet ¡Word Characteristics − Familiar words − Long words 2/7

Components of Speech Communication System Message ¡Context Features − Sentence: meaningful>nonsense − Set size:

Components of Speech Communication System Message ¡Context Features − Sentence: meaningful>nonsense − Set size: 字多<字少 − Guidelines ¡用較少的字 ¡Standard sentence ¡Avoid short word ¡Familiarize user 3/7 Fig 7 -3

Components of Speech Communication System Transmission System ¡Filtering (Frequency distortion) − High-pass: cutoff< 600

Components of Speech Communication System Transmission System ¡Filtering (Frequency distortion) − High-pass: cutoff< 600 Hz − Low-pass: cutoff> 4000 Hz Fig 7 -4 ¡Amplitude Distortion Fig 7 -5 7 -6 − Peak clipping Quality , Intelligibility ≈ − Center clipping Intelligibility − 提高 Intelligibility: Peak clipping Amplify ( 子音/母音 ) 4/7

Components of Speech Communication System Noise ¡Articulation Index (AI) Fig 7 -7 − 1/3

Components of Speech Communication System Noise ¡Articulation Index (AI) Fig 7 -7 − 1/3 octave, S-N, weighted sum − Intelligibility Fig 7 -8 Tab 7 -1 ¡Preferred-Octave Speech Interference Level (PSIL) − Mean of 500, 1000, 2000 Hz (octave) − SIL: Mean of 600 -1200, 1200 -2400, . . . − Intelligibility (vs. distance) Fig 7 -9 − Subjective rating Fig 7 -10 Tab 7 -2 5/7

Components of Speech Communication System Noise ¡Preferred Noise Criterion Curve (PNC) Fig 7 -11

Components of Speech Communication System Noise ¡Preferred Noise Criterion Curve (PNC) Fig 7 -11 Tab 7 -3 ¡Reverberation Fig 7 -12 − Reverberation time: Decay 60 d. B − Reverberation time Intelligibility 6/7

Components of Speech Communication System Hearer ¡Age Fig 7 -13 ¡Wearing of Hearing Protection

Components of Speech Communication System Hearer ¡Age Fig 7 -13 ¡Wearing of Hearing Protection 7/7

Synthesized Speech 種類 Uses Performance Preference Guidelines

Synthesized Speech 種類 Uses Performance Preference Guidelines

Synthesized Speech 種類 ¡Synthesis by Analysis − Digitized human speech compressed data format −

Synthesized Speech 種類 ¡Synthesis by Analysis − Digitized human speech compressed data format − 缺點: 限於 encoded & stored Lack of coarticulation ¡Synthesis by Rule − 缺點: quality 較差