Analyzing the Speech Signal Julia Hirschberg CS 6998 11/28/2018

Basic Acoustics What is sound? Pressure fluctuations in the air caused by a musical instrument, a car horn, a voice Cause eardrum to move Auditory system translates into neural impulses Brain interprets as sound How does it travel? Via sound wave of air molecules that ‘travels’ thru air 11/28/2018

Molecules don’t travel but pressure fluctuations do But sound waves lose energy as they travel --it takes energy to move those molecules And molecules also move for reasons other than e.g. the sound of my voice: noise Ratio of speech-generated molecular motion to other motion: signal-to-noise ratio 11/28/2018

Types of Sound: Periodic Waves Simple Periodic Waves (sine waves) defined by Frequency: how often does pattern repeat per time unit Cycle: one repetition Period: duration of cycle Frequency=# cycles per time unit, e.g. Frequency in Hz=1sec/period_in_sec Horizontal axis of waveform Amplitude: peak deviation of pressure from normal atmospheric pressure 11/28/2018

Phase: timing of waveform relative to a reference point Complex periodic waves (eg) Cyclic but composed of two or more sine waves Fundamental frequency (F0): rate at which largest pattern repeats (also GCD of component freqs) Components not always easily identifiable: power spectrum graphs amplitude vs. frequency 11/28/2018

Fourier’s Theorem Any complex waveform can be analyzed into a set of sine waves with their own frequencies, amplitudes, and phases Fourier analysis produces power spectrum from complex periodic wave Potential problems: Assumes infinite waveform when we have only a small window for analysis Waveform itself may be inaccurately represented 11/28/2018

Types of Sound: Aperiodic Waves Waveforms with random or non-repeating patterns (eg) Random aperiodic waveforms: white noise Flat spectrum: equal amplitude for all frequency components Transients: sudden bursts of pressure (clicks, pops, door slams) Waveform shows a single impulse Fourier analysis shows a flat spectrum 11/28/2018

Sample Analyses Wavesurfer Download from http://www.speech.kth.se/wavesurfer/download.html 11/28/2018

Filters Acoustic filters block out certain frequencies of sounds Low-pass filter blocks high frequency components of a waveform High-pass filter blocks low frequencies Reject band (what to block) vs. pass band (what to let through) 11/28/2018

Production of Speech Voiced and voiceless sounds Vocal fold vibration produces complex periodic waveform Cycles per sec of lowest frequency component of signal = fundamental frequency (F0) Fourier analysis yields power spectrum with component frequencies and amplitudes F0 is first (lowest frequency) peak Harmonics are resonances of vocal folds multiples of F0 Vocal tract filters simple voicing waveform to create complex wave 11/28/2018

Digital Signal Processing Analog devices store and analyze continuous air pressure variations (speech) as a continuous signal Digital devices (e.g. computers) first convert continuous signals into discrete signals (A-to-D conversion) Sampling: how many time points in the signal to consider? Quantization: how accurately do we want to measure amplitude at sampling points? 11/28/2018

Sampling Sampling rate: how often do we need to sample? At least 2 samples per cycle to capture periodicity of a waveform component at a given frequency 100 Hz waveform needs 200 samples per sec Nyquist frequency: highest-frequency component captured with a given sampling rate (half the sampling rate) 11/28/2018

Samping/storage tradeoff Human hearing: 20K top frequency But do we really need to store 40K samples per second of speech? Telephone speech: 300-4K Hz (8K sampling) But fricatives have energy above 4K 16-22K usually good enough 11/28/2018

Sampling Errors Aliasing: Signal’s frequency higher than half the sampling rate Solutions: Increase the sampling rate Filter out frequencies above half the sampling rate (anti-aliasing filter) 11/28/2018

Quantization Measuring the amplitude at sampling points: what resolution to choose? Integer representation 8, 12 or 16 bits per sample Noise due to quantization steps avoided by higher resolution but requires more storage Choice depends on what kind of analysis to be done 11/28/2018

But clipping occurs when input volume is greater than range representable in digitized waveform  transients 11/28/2018

Perception of Pitch Auditory system’s perception of pitch is non-linear Sounds at lower frequencies with same difference in absolute frequency sound more different than those at higher frequencies Bark scale (Zwicker) models perceived difference 11/28/2018

Pitch-Tracking Autocorrelation techniques Goal: Estimate F0 over time as fn of vocal fold vibration A periodic waveform is correlated with itself One period looks much like another (eg) Find the period by finding the ‘lag’ (offset) between two windows on the signal for which the correlation of the windows is highest Lag duration (T) is 1 period of waveform Inverse is F0 (1/T) 11/28/2018

Halving: shortest lag calculated is too long (underestimate pitch) Errors: Halving: shortest lag calculated is too long (underestimate pitch) Doubling: shortest lag too short (overestimate pitch) 11/28/2018

Pitch Track Headers version 1 type_code 4 frequency 12000.000000 samples 160768 start_time 0.000000 end_time 13.397333 bandwidth 6000.000000 dimensions 1 maximum 9660.000000 minimum -17384.000000 time Sat Nov 2 15:55:50 1991 operation record: padding xxxxxxxxxxxx 11/28/2018

Pitch Track Data F0 Pvoicing Energy A/C Score 147.896 1 2154.07 0.902643 140.894 1 1544.93 0.967008 138.05 1 1080.55 0.92588 130.399 1 745.262 0.595265 0 0 567.153 0.504029 0 0 638.037 0.222939 0 0 670.936 0.370024 0 0 790.751 0.357141 141.215 1 1281.1 0.904345 11/28/2018

RMS Amplitude Energy closely correlated experimentally with perceived loudness For each window, square the amplitude values of the samples, take their mean, and take the root of that mean What size window? Longer windows produce smoother amplitude traces but miss sudden acoustic events 11/28/2018

Perception of Loudness Non-linear: Described in sones or decibels (dB) Differences in soft sounds more salient than loud Intensity proportional to square of amplitude so…intensity of sound with pressure x vs. reference sound with pressure r = x2/r2 bel: base 10 log of ratio decibel: 10 bels dB = 10log10 (x2/r2) Absolute (20 Pa, lowest audible pressure fluctuation of 1000 Hz tone) or typical threshold level for tone at frequency 11/28/2018

Pressure of Common Sounds Event Pressure Db Absolute 20 0 Whisper 200 20 Quiet office 2K 40 Conversation 20K 60 Bus 200K 80 Subway 2M 100 Thunder 20M 120 *DAMAGE* 200M 140 11/28/2018

Speech Analysis Gives us Information About variation in Loudness Pitch (contours, accent, phrasing, range) Timing (rate, pauses) Style (articulation, disfluencies) This can be correlated with other features Syntax, semantics, discourse context, words 11/28/2018

Now and Next Week Now: turn in discussion questions and project ideas Read HLT96 (Ch. 5) Try out some TTS systems; exercises Bring 3 discussion questions to class Decide which week you would like to help with class 11/28/2018

Vocal fold vibration [UCLA Phonetics Lab demo] 11/28/2018

Places of articulation alveolar post-alveolar/palatal dental velar uvular labial pharyngeal laryngeal/glottal 11/28/2018 http://www.chass.utoronto.ca/~danhall/phonetics/sammy.html