Hello Fons, Thanks again for your help. I'm good at electronics and hearing science, but Ambisonics and IRs are new to me. It was never mentioned once while I was in grad school (hearing science), so this has been a self-teaching experience. I've received a great deal of help from you and other on the sursound list. As I've mentioned in my past ramblings, stimuli we use in assessing hearing aid or cochlear implant efficacy are far from real world. Simple (but rarely spoken) speech phrases with monaural white noise* maskers provided a convenient yardstick for lab experiments. If a proposed experiment required more than MATLAB and a laptop's built-in sound card, it was frowned upon. I'm an "old school" hardware junkie, but also use MATLAB, Visual Fortran, C, etc. And I try to integrate this background with current issues in hearing disorders. *Note: A lot of people were using white noise (occasionally pink noise) as masking noise. I had suggested the change to speech noise, and did the final mastering and calibration of a now-popular test CD. Some of the original tracks (various sources) were meant to be presented at such-and-such dB hearing level (HL); whereas others were to be calibrated in dBA or dBC. Talk about confusion for the audiologists! I adjusted all tracks to match a cal sound that was to be measured in dBC. Whether one used speech babble or weighted noise tracks, there was no confusion regarding the presentation levels or SNRs. Anyway...
The IRs have been uploaded for you. These particular files were downloaded from a site that no longer exist, so I can't provide the link. The now-missing site was a bit like OpenAirLib.net, but included a classroom, "Great Hall" and "Octagon". The original files were already split into W, X, Y, and Z files (i.e., not interleaved). I converted to 16 bit and 48 kHz to match my existing dry speech (which I recorded). Sony Sound Forge Pro's Acoustic Mirror doesn't require that the bit or sample rates of the IR and wav files match, but I had already made conversions. I'm still processing my own sweeps, which are probably even noisier. My tests were made using a KRK 9000 (I found this to have low distortion across frequencies and uniform response) and a TetraMic. If I remember correctly, the classroom files were made using a Soundfield mic (ST250, ST450, or similar) and a high-end Genelec monitor. You can find the 16-bit, 48 kHz files in the ir folder: www.elcaudio.com/demos/ir/ When using quality IRs from Trillium Lane or Waves, it's easy to change wet/dry ratio to taste. But this is subjective. When using IRs for auralization, I assume 100 percent wet is the norm. In the end, shouldn't the location of the talker (or source) be commensurate with the mic-to-speaker orientation of a given Ambisonic IR? That much isn't too bad, but all rooms appear far more reverberant than what their respective RT60s suggest. And recordings of speech made in reverberant rooms (actual recordings, not processed) sound nowhere as reverberant as processed recordings made with IRs. This is why I ask about the wet/dry ratio. But if empirical or subjective impressions are used, I can't vouch for any scientific validity of the stimuli -- I simply chose it because it "sounded right" (this ain't gonna fly). Many thanks for your time. Best wishes, Eric ________________________________ From: Fons Adriaensen <f...@linuxaudio.org> To: Eric Carmichel <e...@elcaudio.com> Sent: Sunday, December 16, 2012 4:13 AM Subject: Re: [Sursound] Plate Reverb rocks (going off-list) On Sat, Dec 15, 2012 at 03:40:32PM -0800, Eric Carmichel wrote: > One likely reason I was hearing so much "gunshot" noise in my > original samples is because there was other noise in the recording. Yes. The original speech recording contains quite some VLF noise. It could be a good idea to remove that *before* editing, even more so if you do hard edits at zero-crossing points instead of short crossfades. When VLF noise is present the zero-crossings are not those of the waveform of interest and you could end up splicing two different 'DC' levels together. I also suspect there is considerable VLF noise in the reverb IR, it's a typical problem with IRs made using tetrahedral mics. Also this should be removed before performing the convolution. (that's one of the reasons I asked to see the IR) > Clearly, arbitrary starting points aren't arbitrary when it comes > to creating stimuli. Or editing e.g. classical music recordings (which tend to have quite some reverb on them). Here the problem is usually the inverse one: you can't e.g. cut at the start of a take (which is not the start of the piece) as you don't have the reverb from the previous notes. I'm still puzzled by the 'car passing by' sound after the reverberated noise burst. It could indicate something odd in the reverb IR, which is the second reason why I was curious to see it. Ciao, -- FA A world of exhaustive, reliable metadata would be an utopia. It's also a pipe-dream, founded on self-delusion, nerd hubris and hysterically inflated market opportunities. (Cory Doctorow) -------------- next part -------------- An HTML attachment was scrubbed... URL: <https://mail.music.vt.edu/mailman/private/sursound/attachments/20121216/28cde1dd/attachment.html> _______________________________________________ Sursound mailing list Sursound@music.vt.edu https://mail.music.vt.edu/mailman/listinfo/sursound
