Splitting Phonemes Adam Burkhalter Duong Chau Micah Lee
Project Overview ●Isolate phonemes in a signal o Read in an audio file o Separate the individual phonemes o Determine phoneme signature o Output results
Audio Reader Application ●Audio Recorder using Java o Save audio as.WAV o Save audio as bytes ●Playback audio ●Add saved audio
Key Algorithm Vocabulary ●Phonemes ●Exponential Sinusoidal Model ●Accumulated Autocorrelated Functions ●Transients ●Variable Segmentation Strategy
Phonemes Is a basic unit of a language’s phonology, which is combined with other phonemes to form meaningful units such as words or morphemes. Cow has 2 phonemes: \c and \ou
Example of Phonemes /A/a (table), a_e (bake), ai (train), ay (say) /a/a (flat) /b/b (ball) /k/c (cake), k (key), ck (back) /d/d (door) /E/e (me), ee (feet), ea (leap), y (baby)
Transients ●High amplitude, short-duration sound ●At beginning of a waveform ●Occurs in phenomena such as musical sounds, noises or speech
Phoneme Boundaries ●Transients model phoneme boundaries
Exponential Sinusoidal Model (ESM) ●Amplifies signal content o Extreme values get amplified o Low values do not ●Helps determine phoneme boundaries
Accumulated Autocorrelated Functions (AACF) ●Determines the coefficients of the ESM
Variable Segmentation Strategy ●Uses high difference transients to determine phoneme boundaries ●Boundaries are used to isolate phonemes for processing
Phoneme Processing ●FFT analysis to determine signal composition ●FFT data used to create filter banks ●Filter banks used to identify phonemes ●Identified phonemes are characterized ●Results displayed
References Parameterizing Speech Phonemes by Exponential Sinusoidal Model -Jayanth Kumar Talasila and Pradip Sircar