KnowBrainer Speech Recognition
Decrease font size
Increase font size
Topic Title: Microphone Shoot-out
Topic Summary:
Created On: 12/23/2020 02:35 PM
Status: Post and Reply
Linear : Threading : Single : Branch
 Microphone Shoot-out   - ax - 12/23/2020 02:35 PM  
 Microphone Shoot-out   - jmcvay - 12/24/2020 03:42 PM  
 Microphone Shoot-out   - ax - 12/27/2020 02:54 PM  
 Microphone Shoot-out   - ax - 12/27/2020 04:10 PM  
 12/23/2020 02:35 PM
User is offline View Users Profile Print this message

Author Icon
Top-Tier Member

Posts: 433
Joined: 03/22/2012

Well to mark the return to this board, I can't just spam one topic.  And why is there a dearth of Mike information in this place, compared to, say 2014?


I suspect it is in part due to the increase in accuracy of Dragon across the board, which rendered microphone debates less relevant.


Well with my non-optional migration to DME at workplace (embedded version with even less installation than DMO) over the past year, I have been able to do some Mike comparisons.


One only gets a single DME session at one time, which can roam from workstation to workstation (this is a nifty browser feature not contingent upon any VDI).  But one could run a "testing session" simultaneously on a different machine.  Apparently both sessions share the same Profile, which is of course cloud-based.  Aided by the testing session, I have been able to compare mics side by side, two at a time.


First place is SMP 3700/4000, which at "close range", in my LS05 gooseneck converted table mic style, beats out PowerMic 3 just by a smidgeon (as Lunis would say).  However, if speaking from a fair distance (more than 6 inches roughly), PowerMic is slightly better.


I didn't test either directly against my AT2005 in the office as the latter is less "luggable".  However, my subjective impression is that AT2005USB is every bit as good, if not a tiny bit better, at a close distance and in a relatively quiet environment.


All other tested headsets and gooseneck were objectively inferior.

 12/24/2020 03:42 PM
User is offline View Users Profile Print this message

Author Icon
Senior Member

Posts: 154
Joined: 01/02/2015

You may be right about the increased accuracy of Dragon rendering the microphone question somewhat moot. Before COVID hit, I paid an extended on a family member in another state. I took my desktop computer with me in case I got a translation job, but I forgot my SpeechWare 3-in-1 microphone. I did get a translation job and decided to try dictating the translation using the microphone built into my desktop. To my great surprise, it worked almost as well as my Speechware microphone.


James McVay

 12/27/2020 02:54 PM
User is offline View Users Profile Print this message

Author Icon
Top-Tier Member

Posts: 433
Joined: 03/22/2012

Allow me to flesh out my own experience a bit.


During the past year of institution-wide adoption of "front-end" DME, many of my colleagues have commented on the general effectiveness of using "array microphones" built into their Suface Pro, and various off-the-shelf bluetooth devices.


Therefore the general consensus indeed is that recognition capabilities have improved across the board by both Dragon and MModal.  Google and Apple and Cortana etc etc are not far behind, as I could testify from speaking into the remote on my Sony "smart TV" just this week (another lock down in Ontario, might as well replace the TV we have had for 14+ years).


But I quickly turn off the voice function on the TV ... too much progress ... too fast.


All of the above is not to say, however, that as a tool of making a living, microphones are "unimportant".  At most one can say they are a bit less "make-or-break".  Colleagues who are cross-appointed to the the Epic shop in the other end of the town, do comment that the PowerMic Mobile over their phones (for DMO) seem less accurate than the PowerMic at the hospital for DME.


And then there are the perennial speaker factors.  My own limitations are such that I could never make any Bluetooth headset work for me, even the very best Senheiser MB Pro 2 UC I got from Knowbrainer.  The drop in accuracy on DME when I use MB Pro 2 UC is palpable.


Back then it was Lindsay (Monkey8) who commented that "don't expect Dragon to understand you" if "you can't understand you".  By that he simply meant to record a few of your own readings and play back on the PC to hear the differences.  And work to correct some "bad habits".


Now with improvement in AI this and that and better signal processing hardware, the day is probably already here that even when "you can't understand you very well", machine can understand you better.


But nothing replaces objective testing.


One againsts one, I tested the following headsets (otherwise noted, I am the only tester):


Least effective: Bluetooh headset (MB Pro 2).  I hasten to add that I have colleagues who swear by less expensive bluetooth headsets.  But these are the ones who are speaking 6 rooms down the hallway and you have no idea which room they are in but can hear everything they say.


Next up: wired headsets: The old cheapo DMPE 2.x complimentary headset with its USB adapter remains amazingly accurate, paraxodically (subjectively) close to the most accurate wired headset I have used.  However, I didn't compare it directly against PowerMic in DME.  Can't remember why not.  I still have it.  I think the wearing comfort (discomfort, rather) of the DMPE complementary headset is a major strike against it.


I bought both the Senheiser SC 260 (Knowbrainer at the time only carried the monoaural version) and Jabra BIZ 2400 II (monoaural and not carried by knowbrainer, largely on account of some transcriptionist website in the UK swearing by its goodness) in early 2020.  Both were comfortable to use, but the Jabra is a bit ahead in recognition accuracy objectively.  The SC 260 was palpably worse than PowerMic for DME.  BIZ 2400 II was just a "smidgeon" worse than PowerMic.  Biz 2400 II is the only headset that I had a colleague test in addition to myself.  Same conclusion: tiny bit worse than PowerMic at relatively close range.  But the said colleague "won't hesitate" to use it given how close it has performed against this "gold standard".


I also sourced a VEC SoundTech GN-USB-2 gooseneck before the summer.  Knowbrainer carries that as well - as a more affordable impression of "TableMike".  However, by then cross-border shipping had slowed to a crawl, so pretty much only the likes of Amazon with local warehouse were viable options.  And the first one I got from Amazon was actually lost to delivery.  I generally advocate getting microphone as much as possible from Knowbrainer on account of the support one can expect from this place.


Anyhow, the SoundTech GN-USB-2 is not bad, but subjectively not as good as PowerMic and my AT2005USB in the office.


Will continue ... only because it is perhpas not good form to make any given post too long.

 12/27/2020 04:10 PM
User is offline View Users Profile Print this message

Author Icon
Top-Tier Member

Posts: 433
Joined: 03/22/2012

Just before the pandemic hit, I contacted Philips to request to test out their SpeechMike range of products.  They actually put me in touch with their AD (or AD for "ADs") in Eastern Canada.


As a result I received both an SMP4000 with AirBridge, and a SpeechOne PSM6300 (yes that one).  


First off for the entire 8+ months I had them, I never so much as put on the SpeechOne.  I took the headset out of the plastic and "beheld" it.  And that was that.  By then the plague had already engulfed the land.  I reckoned there was no way I was going to pass around something with soft rubber for going on ears to colleagues for testing, even if we were going to virox it.  And no-one knew how much viroxing this thing could withstand.  And I wasn't sure how much virox I would want to drip down my inner ear ... you get the thought process.


And the size of that thing ... do I really want to look like I am wearing a Starbucks "what's your drink today" headset?  I am pretty cavalier already around patients (perhaps more so than I should).  But there are some limits even I hesitate to "transgress".


Then there is the price of that thing.  I only wanted to test it, and not buy it unless it "sang to me".  So I sent it back last month completely unused.


The SMP4000 + AirBridge, however, is a very capable combo.  Many of the physician regulars here already know that and have been using them for a couple years at least.  Perhaps half a dozen colleagues of mine tested it over a period of months and all agree that the SpeechMike Air is subjectively indistinguishable (will expand on that) from PowerMic in terms of accuracy under DME.  


As I returned the SMP4000/AirBridge (I did have to pay shipping both ways) to the AD, I ordered an SMP3700 for myself.


The reason I chose the wired SpeechMike is simply that it fits my style better.  I almost exclusively use it as a Table Mic sitting on the Lamicall LS05 stand.  Furthermore, when EMP shocks arrive in 2021, during a "solar flare" (not the "SunBurst" currently making rounds ... or maybe that, too), I could perhaps dictate a few minutes longer than those using the Air.


I would personally pick the Air with AirBridge if I were using a laptop primarily.  In that scenario the wirelessness makes a world of difference.  And not one colleague, or myself, has noted any hitch at all with the AirBridge.  I had colleagues in ER test it and it was no problem.  Mind you, however, that it was the only SMP4000 in our ER at the time.  Wouldn't know how multiple SpeechMike Air in close proximity operate in "complex EM environment" (such as ER) in real life.


It was THAT good.  However, once Uncle Sam upgrades GPS to Next-Gen or rolls out 6G/7G, would the Air still be good?  Only time will tell.


I am a (forced) convert to "defensive computing".  So when it comes to forking out of my own pocket, I chose the more conservative SMP3700.


By the way, for those who think a USB chord "sitting there" just "can't break" (count me as one of them - at least used to), I had the OEM long USB cord that came with my AT2005USB fail without warning.  It was always on the OEM swing stand so there was no picking it up by hand, shaking it, stretching the cord, jigging the connector, or any of that sort of shenanigan.  It just failed this year for no reason.  A Google search on AT2005USB shows that it is a "common problem" with this package.


And that's from Audio Technica, a reputable company.  I now got rid of the swing stand and use the Lamicall LS05 with the AT2005 as well.  Much better stand any way, with a new USB cord.


Given the doubling of copper price and further hikes next year, how robust do you think the USB cords you buy in 2021 will be?


Final comments on PowerMic III vs SMP3700.  Based on my single-person objective testing, SMP3700 beats out PowerMic III in terms of accuracy in usual hand-held distance, in a quiet environment, used for DME.  I didn't test their performances in a noisy environ.  One, it takes time.  Two, there is one kind of "quiet", but many different kinds of noise.


The difference is small, however.  To give one example:


SpeechMike: "endoscopic sleeve gastric antral plasty"


PowerMic: "endoscopic sleeve cancer trial plastic"  


Both were wrong as the word was "endoscopic sleeve gastroplasty".  But SpeechMike was just "less wrong".  This is of course with reading quite quickly.


Over 8 or 10 inches away or more (previously I said 6+ ... but I think it is further), PowerMic III performs better, perhaps due to some unique "far field" signal processing.  The difference here is larger than the difference between the 2 in usual hand-held distance (reiterate, PowerMic is better in far field).  But for medical/work use, I can't think of one scenario where someone would prefer "far field" when "usual distance" is available.


Last thing: there are so many Zoom/Teams meetings these days as to make folks want to puke.  But the speaker on SpeechMike is actually not half-bad for one of those meetings.  PowerMic just processes sound differently I think both at the Mic level and speaker level.  "Tape" your own speech with PowerMic and you will hear.  It records more "harshly" and plays more harshly (even clips recorded by SpeechMike or others).  Using a PowerMic for Zoom/Teams would be simultaneous torture of yourself and your colleagues.

KnowBrainer Speech Recognition » Microphones and Sound Cards » Microphone Shoot-out

32325 users are registered to the KnowBrainer Speech Recognition forum.
There are currently 1 users logged in.
The most users ever online was 12124 on 09/09/2020 at 04:59 AM.
There are currently 491 guests browsing this forum, which makes a total of 492 users using this forum.

FuseTalk Standard Edition v4.0 - © 1999-2022 FuseTalk™ Inc. All rights reserved.