Lunis,

I bought a Sony SX25 1-1/2 years ago, after comparing it with Olympus recorders mainly on the basis of their performance with NaturallySpeaking in Nuance's in-house testing:

- Sony SX25: 5 dragons

- Olympus DS-4000: 2 dragons

It seems to me that Nuance's in-house testing is both unbiased and authoritative.

Also, the Sony's best sampling rate is 44.1KHz, while the Olympus's best is only 12 KHz.

Finally, the Sony's best frequency response is 60Hz - 13.5KHz, while the Olympus's best is only 300Hz - 5KHz.

Recently I noticed that you sell only Olympus recorders. If voice transcription weren't a factor, I can see why you'd recommend them over the Sony recorders. But given the Nuance test results and the substantial difference in the sampling rates and frequency responses, plus the fact that your primary market (I presume) is users who want to do voice transcription, I'd appreciate if you could tell me why you recommend the Olympus recorders over Sony (in case there'd be a good reason for me to replace my Sony with an Olympus).

Thanks.

RMB,

Nuance test results for microphones are mostly polar opposite of our tests. They seem to test for pure audio throughput and not real life people to computer dictation. There are many other factors that account for accuracy. Among these are noise canceling and echo rejection. To achieve these desirable qualities the sound in great speech recognition microphones is somewhat muffled and thus do poorly in Nuance's testing. I suspect the same with their recorder tests.

The fact that the Olympus has a narrower frequency response (closer to the range of the human voice) may explain why the Olympus performs better than the Sony.

Martin

Martin Markoe - BANNED USER: This user has been banned from these forums

Do yourself a favor and take Marty's advice.  The reason for it is this.  It doesn't matter that Sony's overall sampling rate is higher than Olympus's.  The reason for this is simply that DNS only accepts 11.025 kHz 16 bit Mono sampling rate, which translates to a maximum frequency of 5 kHz.  Even if you import a wav or MP3 file for transcription that is 16-bit Mono but may range from 11.025 kHz to 99 kHz, DNS extracts only what it needs to provide 11.025 kHz 16 bit Mono sampling for transcription.

Therefore, if you want to use your Sony as an MP3 player, fine.  However, keep in mind that with regard to speech recognition, the higher the sampling rate the greater the propensity to introduce noise.  Noise is an accuracy killer.  This is not to say that the Sony recorders aren't good, it's simply saying that they are overkill and the Olympus is better (i.e., more accurate) for the reasons that Marty's state's.

In addition, I'll add one further comment.  Somewhere along the line during the acquisition of all of the L&H resources, which included all of the Dragon Systems facilities and equipment, ScanSoft, and subsequently now Nuance, either divested themselves of or lost their primary testing equipment which was originally designed by Labtec, now Logitech, called the “MOUTH.”  During my tenure at L&H, and after the acquisition of Dragon Systems and DNS, we used the “MOUTH” for testing.  During that period we turned the entire compatibility chart upside down.  Our testing was more in line with the results that Marty gets.  However, ScanSoft/Nuance, for what ever reason, decided to throw out all of that research that we did.  What they show in their compatibility matrix is really, mostly, old data that has never been completely or effectively updated.  If you base your purchasing decisions on that information, I have a couple of bridges in New York City that I'd like to sell you.  ;-)

Chuck Runquist
Former DNS SDK & Senior Technical Solutions PM for DNS

If you hear the sound of hoofbeats, think horses not zebras.
Law of Parsimony (Occam's razor)

Chuck,

> During my tenure at L&H, and after the acquisition of Dragon Systems and DNS, we used the “MOUTH” for testing. 
> During that period we turned the entire compatibility chart upside down.  Our testing was more in line with the results that Marty gets. 
>  However, ScanSoft/Nuance, for what ever reason, decided to throw out all of that research that we did.  What they show in their
> compatibility matrix is really, mostly, old data that has never been completely or effectively updated. 

Thanks for that information. Never heard that before. I always thought "The Mouth" was my boss. I hope she does not read this. :-)

 Marty

Martin Markoe - BANNED USER: This user has been banned from these forums

Click KB 2012 REV D to Download a 30 Day Evaluation of KnowBrainer 2012 

Thanks to all for your replies.

I can now see, based on Chuck's info regarding the specs, why "on paper," at least, the Olympus should perform better than the Sony.

But I'm particularly interested in the "bottom line": Martin said that the Olympus performs better than the Sony, and Chuck concurred. Can either of you tell me whether this is based on not just the "on paper" specs but on an actual hands-on comparison? And if so, did the comparison result in quantitative test results, or was it just a subjective impression?  (Ideally, I'd like to know how many percentage points my accuracy might improve by moving from a Sony to an Olympus. For example, I know from reading posts from at least one of you that the best that can be achieved with a digital recorder is about 2% less than that with a high-quality microphone.)

Lunis: on your product page for the DS 3300, you say "Professional digital recorder utilizing the same speech algorithms and recording media as a DS 4000. As far as speech recognition quality is concerned ... [t]he DS 4000 and DS 3300 are identical." What about the DS-2300? Can you tell me in what way, and/or to what extent, its speech recognition quality is inferior to that of the DS 4000 and DS 3300?

Thanks.

www.itspeaking.co.uk