Here in the year 2001, it’s too bad we can’t talk to our computers in the same easy, informal way Dave addresses the HAL 9000 in
2001: A Space Odyssey.
Still, speech is one of the most progressive, futuristic methods of bringing data into a computer, and it’s beginning to play a prominent role in voice recognition, Internet telephony and broadcasting, video conferencing, and voice chat.
To engage in any of these activities, however, you need to buy a USB microphone, since Macs no longer come with mics and use USB for all sound input. Fortunately, there’s no shortage of options. We rounded up ten head-worn mics and one desktop mic and put them through their paces with voice-recognition and recording software. The Plantronics DSP-400 and DSP-500 were the best performers overall, producing excellent recordings and offering the most amenities, while the Telex M-560 and Andrea Electronics NC-7100 performed the worst with speech-recognition and recording software — which is surprising, since the latter model ships with both
Meet the Mics
When you plug one of these mics into a Mac running OS 9, the option to use a USB mic appears in the Sound control panel, so no drivers are required. (In OS X, the Listening control panel — which governs voice-command preferences — assumes a USB mic as the default.) The Plantronics mics come with Mac drivers, which allow you to modify the mics’ firmware. You’ll have to run one such patch for the mics to work with speech-recognition software, such as IBM’s ViaVoice or MacSpeech’s iListen. Thankfully, the process is quick and painless and doesn’t require a restart.
All the mics use USB exclusively except the Parrott Jamaica and the NC-7100, which use a one-eighth-inch audio jack for the speaker. If you have a newer Mac that lacks audio jacks, you’ll want to go with a different mic.
Most of the mics we tested are head-worn “boom” mics — that is, the mic itself hovers near the corner of your mouth on a stem that extends from the headband. The sole exception is the M-560; this long, thin mic sits on your desk, aimed at your mouth, and attempts to isolate your voice from the surrounding sounds.
Most of the other mics come with speakers built into the headband, which is a useful addition. (The Telex H-531 and H-831 lack speakers. Don’t be fooled by the Telex H-531’s foam speaker cover — there’s nothing inside it.) With speakers, you can monitor your mic’s sound quality, as well as talk with others over the Internet without disrupting those around you. Also, these headsets’ foam-covered speakers make them more comfortable to wear than those without speakers. The NC-7100, the Plantronics DSP-100, and the VXI Parrott Aruba and Parrott Jamaica have a single speaker on the left side of the headband and a clip on the right, to keep the headset in position.
In contrast, the Parrott Caribbean and the Plantronics DSP-300, DSP-400, and DSP-500 all have stereo speakers. Those of the DSP-500 are especially large, light, and comfortable, fully encasing your ears in foam. The stereo speakers sounded better than the monophonic ones, but I preferred the DSP-400’s speakers even over the larger speakers of the DSP-500 — they packed a bit more punch.
Stereo speakers not only make mics more comfortable to wear but also make it easier to adjust the boom, because you already know where to put the earpieces. The monophonic (and speakerless) head-worn mics require a bit of futzing before you can use them with delicate sound programs, such as speech-recognition software: you have to position the boom so that the mic is just to the left of your mouth, but not touching it. The DSP-500 was the best in this respect: I was able to toss it on and position the mic with a minimum of fuss. The DSP-300 was a bit harder to adjust than the others because it uses a rigid boom that only slides forward and back, for a narrower range of positions.
A handful of these mics have some useful, distinguishing features. Unlike the other mics, the Parrott Aruba and Parrott Caribbean have volume and mute controls in a small unit that you can clip to your shirt (the Plantronics models do, as well, but their controls work only with Wintel boxes). The DSP-400 is unique in that it folds and comes with a vinyl pouch, making it that much more travel-friendly.
Do You Read Me?
I tested each mic with MacSpeech’s iListen 1.1 (which ships with the Parrott Jamaica). Without correcting mistakes — so the program wouldn’t “learn” and create an unfair advantage for mics tested later — I dictated a 94-word passage and counted the errors to see if there were significant differences in the mics’ speech-recognition performance. Then I double-checked the mics’ performance with ViaVoice’s Mic Setup utility.
The Telex H-831 helped iListen perform extremely well — the software made only five errors. With all three of the VXI mics and the H-531, iListen made eight errors or fewer; with the Plantronics mics, it made between 10 and 12 errors, which is still quite good. With the NC-7100, iListen made 16 errors.
The freestanding M-560 didn’t work at all with iListen; the program crashed whenever I tried to run it with the M-560 plugged in. The mic did work with ViaVoice Enhanced Edition, but the program’s Mic Setup utility rated the mic’s sound quality as merely Fair (on a scale of Poor, Fair, Good, Very Good, and Excellent). ViaVoice rated the DSP-400 and 500 as Very Good and all others as Excellent.
For the Record
To evaluate the mics’ ability to faithfully capture a voice with a minimum of noise, I recorded a snippet of speech using each mic and saved the recordings as AIFF files. Then a four-person jury listened to the files and compared their clarity and the amount of white noise they contained.
The DSP-500 produced by far the clearest recordings. The Parrott Aruba, DSP-100, DSP-300, and DSP-400 all produced acceptable, relatively clear sound files. The H-531 and H-831, though they both did very well at speech recognition, produced sound files that were very noisy and muddy. The Parrott Jamaica, Parrott Caribbean, and NC-7100 also produced poor sound files. The M-560, despite its attempts to capture just the voice and bypass the surrounding noise, produced the worst-sounding files, inspiring two jurists to comment that the files sounded as if they were recorded on a freeway.
Many of the mics offer ways to cancel noise, but the presence or absence of such a feature had no effect on the recordings’ overall sound quality. With some files, such as those of the H-531, there was very little noise but the signal was still weak; the DSP-500, although it produced the best-quality recordings, also captured some noise. In the end, we found that the DSP-500 struck the best balance between sensitivity and clarity.