Filters and Speech Acoustics Flashcards
What is a filter?
- An acoustical system that changes the spectrum of a sound
- Isolates and emphasises a particular segment of the frequency system, affecting timbre
How does filtering compare to resonance?
- Filtering implies that there is less of the frequency segment in the output vs the input
- Resonance implies that there is more of the frequency segment in the output vs the input
What is the Natural Frequency (f-nat/fc)?
The frequency corresponding to the maximum amplitude of vibration
What is the Upper Cutoff Frequency (fU)?
- The frequency above fC for which the amplitude of the response is 3dB less than the response at fC
- Acoustic power is halved compared to fC
What is the Lower Cutoff Frequency (fL)?
- The frequency below fC for which the amplitude of the response is 3dB less than the response at fC
- Acoustic power is halved compared to fC
What is Bandwidth (delta f)?
- The range of frequencies passed by the filter
- Bandwidth = fU - fL
- Narrowly tuned filter has smaller bandwidth and vice versa
What is the Attenuation Rate?
- The slope of the filter curve, expressed in dB per octave
- The rate at which frequencies above or below fC are attenuated
- Quantifies the selectivity of the filter
What is a low-pass filter?
Passes energy below fU, attenuates energy above fU
What is a high-pass filter?
Passes energy above fL, attenuates energy below fL
What is a band-pass filter?
- Passes energy between fL and fU, attenuates energy outside of this
- Signal passes through low-pass filter then high-pass filter
What is a band-reject (notch) filter?
- Rejects energy between fL and fU
- Attenuates energy below fU and above fL
- Signal passes through low-pass filter and high-pass filter simultaneously
What is a constant bandwidth filter?
- A type of band-pass filter where bandwidth is independent of fC
- Bandwidth of each filter is identical (white noise input)
What is a constant percentage bandwidth filter?
- A type of band-pass filter where bandwidth is not independent of fC
- Bandwidth is always some constant percentage of fC
- Octave filter: bandwidth is always 70.7% of fC (0.707 x fC)
Describe the output of an octave band filter
- Output of the octave filter depends on fC
- Doubling fC doubles the width of the octave filter
- The filter passes double the intensity, resulting in a 3dB increase in output (when filtering white noise)
Describe white noise
- Equal energy per frequency, resulting in a consistent and uniform sound spectrum
- Perceptually, more of mid-high frequencies are audible due to amplification of mid frequency sounds by middle ear
- 0dB/octave slope
Describe pink noise
- Equal energy per octave
- Spectrally, more energy in the low frequencies compared to white noise
- Perceptually, more of low frequencies audible, sounds lower in pitch than white noise
- Minus 3dB per octave
What happens when correlated waveforms are added together?
- If same phase = double in sound pressure = 6dB increase
- If opposite phases = phase cancellation = sound pressure of 0
How do complex period waveforms occur?
Pure tones (periodic waveforms) of different frequencies are added together
What are harmonics?
- The sinusoidal components of a complex periodic wave
- Have frequencies which are integer multiples of the fundamental frequency (f0)
What are the steps in the source filter model of speech production?
- Initiation from an energy source (e.g. vocal fold vibration, turbulence of air through a constriction)
- Resonance of the sound producer
- Radiation of sound into the air
Describe the waveform of vocal fold vibration
- Complex periodic
- Approximately sawtooth in shape (glottal pulse)
What affects the fundamental frequency of vocal fold vibration?
- Length of vocal cords
- Tension in vocal cords
- Mass of vocal cords
- Pressure in trachea
What is the range of F0 of speech?
60Hz-500Hz
What are the three major contributors to prosody?
- Fundamental frequency (voice pitch)
- Amplitude envelope (vocal effort)
- Duration and rhythm (timing)
What is the significance of quarter wavelength resonators?
- When waveforms reflect back on themself in a tube, it will be filtered/resonated based on the length of the tube
- When the wavelength of a tone is 4x the length of the tube, that frequency will be maximally resonated
- Longer tubes resonate lower frequencies, shorter tubes resonate higher frequencies
What is the first formant (F1) and its equation?
- F1 is the centre frequency of the first of a series of resonances
- Wavelength of F1 is 4x the length of the tube
- F1 = s/4L
- S = 334m/s, L must be in metres
