3-D Spatialization and Localizationand Simulated Surround Sound with Headphones Lucas O’Neil Brendan Cassidy
Overview • 3D with headphones • HRTF Model • Convolution • 360 + elevation Panning • Upmixing • Pro Logic • Delays • Filters • Sub • Autopanning • Downmixing with HRTF
Mathematical HRTF Model • We tell 3D directionality through 3 cues other than just ITD and IID • Pinna Reflections • Shoulder and Torso Reflections • Head Shadow and ITD • Can model using filters/delays.
Shoulder/Torso reflection simulated by echo: • Pinna reflections via a tapped delay line
Head shadow diffracts the sound wave. • Simulated in digital domain by 1st order IIR filter: • ITD due to separation obtained by allpass filter with group delay:
Input azimuth and elevation angle. • Delay by shoulder echo. • Add delay line due to pinna reflections. • Filter through Head Shadow and ITD filters. • spatialization~!
Convolving with HRIR • HRIR = Head Related Impulse Response • Measured with KEMAR dummy (MIT) • Convolve audio with impulse response corresponding to appropriate angle
360 Corkscrew Panning • Demo to shown point source spatialization. • Pick rotation frequency for azimuth and elevation. • Breaks up signal into blocks and performs HRTF with different angles on each block to simulate 360 rotation around head and elevation from -90 to +90 • Done with both mathematical model and convolution techniques.
Upmixing to 5.1 Surround • Investigated Dolby Pro Logic decoder. • Initially used gains/phase shift matrix to split up stereo signal • Tweaked further adding delays to center and surround channels
Pro Logic II has 3 modes of operation: • Movie (not used in this project) • ‘Pro Logic’ • Music • Surround channel uses 7kHz LPF in Pro Logic mode • Surround channel uses Shelving Filter in music mode • Used 4kz cutoff for shelving. • Surround channel has 20ms delay in Pro Logic mode, but not music mode
Subwoofer Simulation • 5 channel surround was losing some low frequency due to cross talk corellation and phase cancellation in surround channels. • Solved by cloning low frequencies of signal (using 300Hz LPF), then mixing them back in the stereo channel after downmixing the 5 channels
Autopanning • Pro Logic mode uses autopanning to detect directionality and adjust 5 speaker mix. • Preserves RMS energy in signal.
Sub-band Autopanning • Scope of project did not allow for implementation • Surround upmixer breaks signals into bands and pans those bands to the appropriate location • Can detect different instruments in music (like frequency keying in DAW software) and localize each instrument.
Resultant stereo sound file has convincing spatialization effects. • Pro Logic mode autopanning implemented without sub-band separation tends to have vocals that jump back and forth across the left and right channels. • Music mode sounded better (for music).