Previous post:

Next post:

Voice Recognition FAQ – Microphone and performance help, information and recommendations

January 1, 2005

in All Articles,Audio,Books and documents

I’ve been using voice recognition software for a long time (see the full review of Dragon NaturallySpeaking and IBM ViaVoice). I often get questions from readers asking about recommendations on hardware or similar issues. Here are some of the most common.

Is there a good microphone you recommend? What microphone setup do you need for quality voice recognition with Dragon NaturallySpeaking?

Generally, you want to get a high-quality microphone, and possibly what is called a USB sound card or USB pod. While the microphones that come with voice recognition software like Dragon NaturallySpeaking or IBM ViaVoice are okay, they could be much better. You’ll get optimal performance by upgrading to a professional-level microphone setup. This is strongly recommended if you’ll be using voice recognition software for any extended length of time.

This is what I use, and is currently the absolute best microphone setup you can get, not just for Dragon NaturallySpeaking, but any speech recognition software package:

VXI USB sound card pod (or get the Andrea USB sound card pod, if you have a quiet speaking voice)

All are available from the good people at Note that they do some physical modifications (all covered under warranty) to make the microphone work properly in this configuration. This is the mic setup I currently use and greatly prefer over the low-end mics that came with the software, both in terms of comfort and performance.

The microphone is usually a unit that is cabled to plug in to standard analog audio jacks in a PC’s sound card. However, depending on your setup (specifically, if you have a good sound card or not), you may need what is called a USB sound card pod. See below.

When I dictate, the Sennheiser microphone is plugged into the VXI pod, which is plugged into my laptop.

What exactly is a USB sound card pod?

A USB sound card pod, or “USB pod”, is just an external sound card that plug into your computer’s USB port. The point of having one is to bypass the existing sound card on your PC, as 1) that sound card may be low-quality, and 2) internal sound cards are more prone to pick up electronic noise when recording (this happens very often in laptops).

Laptops often have below-average sound cards, and require USB pods. Desktop PCs aren’t immune either, though eMicrophones tech support says they usually have pretty good luck with SoundBlaster sound cards.

Note that USB sound card pods do not work well with USB hubs. Connect your USB pod directly to your computer’s USB port.

You recommend both the VXI and Andrea ISB pods. Is there a difference? Is it possible that some types of voice recognition software may work better with one voice type than another type?

Yes, though the differences are small. I’ll let Michael from eMicrophones explain:

“Most people, probably over 90% of users, can use either the Andrea or VXI pods interchangeably. Those people with loud booming voices (me) benefit from a VXI Pod since it has a lower gain. I don’t know why I’m so loud. Maybe it’s from screaming at my kids or my past work, college radio station DJ. Conversely, people with very soft voices may benefit from an Andrea Pod.”

What are the “physical modifications” done to the Sennheiser microphone?

Michael from eMicrophones explains:

“The Sennheiser headset as is, is not meant to be used with a Soundcard or USB pod. It plugs into the Sennheiser wireless transmitter, which is a different kind of connection. One of the modifications has to do with changing this connection so the microphone can plug into a pod. Another has to do with opening the housing and adding special acoustic tape to reduce echoing.”

I’m hearing static when I record audio with my microphone. Dragon NaturallySpeaking’s audio setup wizard is failing me! What kind of results are normal?

When you record your own voice, you should hear no static at all (use this link for testing instructions). If you do hear static, you may need a USB sound card pod. This should clear up the bad reception, which is caused by electronic noise that is picked up by your internal sound card. The USB pod is a way to bypass your internal sound card and greatly increase signal quality. (Symptoms that confirm this problem are that the issue only happens when your laptop is plugged in to AC power. When running on batteries, there is no static. Also, reversing the plug, thus reversing the AC polarity, sometimes helps.)

If you are already using a USB pod, then there may be more significant problems. Contact your pod and microphone supplier for tech support, or use the resources at the end of this page.

As for Dragon NaturallySpeaking’s audio setup wizard, you should run this every time you switch environments or microphone placement, to allow DNS to factor in different background and foreground noise. The audio setup wizard’s “score” is just a measure of your dictation’s signal-to-noise ratio. Any quality microphone should allow you to consistently get a score of 20 or better.

I also use Mozilla’s Firefox / Thunderbird for personal email and web browsing and it doesn’t look like Dragon NaturallySpeaking supports it.

They support third-party programs to a smaller extent. And if you get the “Professional” version, you can create “macros”, which allow you to create your own commands and have those commands interact with any software, including Mozilla’s. And apart from specialized commands, you can still use DNS to dictate to any text window.

How do you get over talking to yourself? Do you work from home, have a single walled office? Does nobody hear you talking to yourself?

When I use voice recognition, yes, I’m at home or in an enclosed office 99% of the time. I did a few tests in public, but to make it work properly I had to talk at a normal voice, and I figured that’d be too annoying to those around me. So yes, when I do this I’m secluded.

Do you work for or are you in any way associated with Dragon Naturally Speaking as a software promoter or any type of financial beneficiary?


I use what is known as a stenomask and I whisper; which software works better for that method of speaking?

That’s a good question, and to get an answer I’d point you to the microphone experts hereor here. Either place you’ll get good tech-level hardware support for issues like mic performance when whispering. However, my own opinion is that it will not work well, certainly not as well as when speaking normally. Voice recognition works very well for people speaking in a natural voice, and wasn’t designed to interpret whispering.

What should the processor speed be on a laptop to work best with voice recognition and what other specs, videocard, soundcard, ram, etc are suggested. Is there anything that is an absolute no-no? The one thing that has been voiced by both Dragon and IBM reps to me is not to go with Gateway.

I’d also recommend to stay away from Gateway (as well as eMachines). I do so at my page where I recommend a good “standard PC”. Use these stats as absolute minimums for a new system, and read the text lower in that page for more detail.

Note that unless you’re playing newer video games or working with multimedia (audio/video editing, etc), video card and sound card choice don’t make much of a difference. I say that even in the case of voice recognition, because for voice recording you should get a USB sound pod, which nixes the sound card requirement:

In your case, I’d recommend bumping the RAM even higher if it’s affordable, like to twice my recommended stats on my “standard PC” page. This is a major help to system performance, particularly when working with specialized, CPU-intensive software like voice recognition.

To specifically answer your questions, Dragon NaturallySpeaking has minimum system specs here. Download and view the PDF of the Dragon NaturallySpeaking edition you’re interested in and check them out. You’ll note that the system requirements for voice recognition are not very high. However, you won’t get very good performance from a system running these specs. Get something more powerful, like the configuration I recommend above.

I speak at least 225 words per minute and I need a speech engine that can keep up with me. My question about that is this: Is that going to depend on the speed of the computer or the quality of the speech software?

Both. But keep in mind Dragon NaturallySpeaking is designed for those, well, speaking naturally! It’s not intended as a high-speed dictation system. Speeds like 150-175 wpm are easily achievable, and faster than that may take more training and/or better vocal cords than I currently have. I am not aware of any high-speed speech-to-text software packages.

One of the options within Dragon NaturallySpeaking is “Speed vs. Accuracy” (available in the Options tool, Miscellaneous tab). I would think you could crank this to the “Most Accurate” setting, and given a fast PC and proper fast but accurate enunciation, you could get good results with fast speaking.

What do you think of Microsoft Speech?

Microsoft Speech, unfortunately, is not something I’ve used or had time to include in my own testing. However, the one place I can point you to is, a good place for Microsoft Speech tips, information and analysis.

I have an nephew who has Muscular Dystrophy and uses IBM Via Voice for dictating papers and email. Once we got a good microphone it’s been satisfactory. However it leaves him tethered to the computer till somebody comes in and takes off his headset. If there was a wireless microphone with high enough quality to use with voice recognition software, it would be quite a blessing for people like him. Are you aware of one?

I’m aware of some for Dragon NaturallySpeaking, with possibilities for IBM ViaVoice.

Go to the Nuance hardware compatibility page.

If you choose to view “Dragon NaturallySpeaking”, you’ll see a list of several recommended wireless microphones for Dragon Naturally Speaking. As DNS is the best voice recognition software you can get, and is price-comparable to ViaVoice, Id’ recommend getting it since you’ll be spending money on a new microphone anyway. However, you can also use the dropdown to view compatible microphones for ViaVoice. You’ll have to browse through that selection manually to see which ones are wireless.


Where can I go for more information, or voice recognition tech support?

For speech recognition and voice recognition technical support, usage techniques and other software and hardware recommendations, see:

This article is Andy’s analysis and review of Dragon Naturally Speaking and IBM ViaVoice.

These articles are a vast collection of speech techniques and technical software and hardware discussions that apply not only to the software mentioned here, but to voice recognition tech in general.

An excellent help forum for Dragon NaturallySpeaking.

The Google Group for voice recognition.

The official Dragon NaturallySpeaking help forum.

1 Focus Medical Software is a reseller of Dragon Naturally Speaking, and can help specifically with DNS medical implementations.

Previous post:

Next post: