e.Digital Corporation (fka EDIG)

[Tinroad]

Don't be so hasty in assuming "VoiceNav has never gotten a single positive review" (especially since you admit to having no experience with it). Here's what one reviewer had to say about it after taking the time to actually give it a workout:

Review: eDigital MXP100 with VoiceNav

By Richard Menta 2/09/02

We have to admit, we took our time with our review of eDigital's latest player. Who wouldn't with an entire gigabyte of songs stored in a flash-sized player. You can thank IBM's Microdrive, a minute hard drive of identical dimensions to compact flash cards that offer capacities of 340MB, 512MB, and 1GB. The MXP 100 uses this drive and leverages it advantages into a competitive MP3 portable that offers more memory than standard 64MB players without ballooning to the size and weight of jukebox portables like the Nomad Jukebox.

It also offers another feature that is actually more practical than gimmicky. Using a technology called VoiceNav, you can jump directly to a specific track by simply saying the songs name. When you have a few hundred songs to scroll through, the ability to jump directly to a song by simply saying its name becomes a welcomed convenience.

The eDigital MXP 100

The Hardware

As we said, the e.Digital MXP 100 comes with a choice of 340MB, 512MB, and 1GB MicroDrives and is about 25% larger than a Rio 500 flash portable. That makes it small enough to conveniently fit in a pocket.

The MXP 100 has 8MB of built-in memory for skip protection and can play tunes compressed in the MP3 and WMA format. Files are transferred to the MXP 100 using its Music Explorer software through a USB connection. Rounding out the features, the MXP 100 has voice recording capabilities, the VoiceNav feature mentioned above, 8MB of built-in memory for skip protection, and runs on an excellent set of lithium-ion batteries that had no trouble handling the power hungry Microdives. We got several hours of use out of them, not as much as the 10 hours Apple claims its iPod gets from its lithium polymer battery, but still good.

The player handles tunes using the MP3 and WMA formats. The MXP 100 is secure music capable supporting several digital rights management protocols including Intertrust, Windows Media DRM and IBM's EMMS.

As of this writing, pricing for the MXP 100 are $399 for the 1GB version, $349 for the 512MB, and $299 for the 340MB. We tested the unit using both the 1GB and 340MB drives. The MXP 100 will also take all sizes of CompactFlash cards including SanDisk's soon to be released 1GB version, which will use considerable less power than the MicroDrive, but will run more than double the price when released.

VoiceNav

What's most interesting about this player is that the unit's most unique feature is not the MicroDrive itself, but a user interface called VoiceNav. VoiceNav allows the user to control the unit using vocal commands. For those of you who have used jukebox or other high capacity portables, you know how tedious it can be search for a particular song when there are hundreds of them to scroll through.

Developed by Bell Labs' Advanced Technologies (a division of Lucent) the MXP 100 is the first consumer product to use this voice navigation interface. Once engaged, the user can navigate playlist folders, select tracks, and play tunes by simply speaking one of several command words into the built-in microphone.

VoiceNav has an fluid phonetic dictionary embedded within it offering speaker-independent recognition. Unlike a software product like IBM's ViaVoice, which needs to learn each individual user's particular speech patterns over time through regular use, VoiceNav requires no such learning input. That's a heck of an accomplishment if it works well.

So, how well does it work?

The basic commands for VoiceNav are Play, Folder, Track, Next, Back and Done. The MXP 100 also recognizes track and folder names, analyzing the spelling of each word, applying standard pronunciation rules and the attempt to match the expected wave patterns of these words with any words spoken into the unit's internal microphone.

Considering the broad palate of individual speech patterns including accents and regional dialects, this is not an easy job and why the progress of voice recognition over the last decade has been a challenging one. Think if President George Bush used this player to call up a track name with the word "Nuclear" in it (he pronounces it "Nu-cu-lar") and you get the idea.

VoiceNav's includes the most up to date applications of speech recognition technology and, as someone who has used IBM's ViaVoice, it shows it. Music Explorer's interface works here too, insuring that track titles and artist or album information from digital music tracks are saved and transferred to the MXP 100 as recognizable phrases.

To activate VoiceNav you press the record button when in music play mode (you need to press and release the button. If you hold it, you will activate the unit's internal voice recorder). Navigation using VoiceNav only operates when a song is not playing (manual controls will allow navigation when a tune is pumping), therefore there is no "Stop" or "Pause" command. The "Done" verbal command is used to disable VoiceNav vocally.

Using the basic commands only, we had no trouble navigating back and forth between folders and songs. We even tried various accents to throw the player off, everything from Brooklyn to southern to bad impersonations of various Monty Python and Simpson's characters. It is very impressive when you shout the word "Folder" talking like Apu Nahasapeemapetilon and it still works.

We could fool it a few times, and sometimes the word "Play" would act a little sluggish, but overall it got off to very good start.

Bearing in mind that speech recognition not yet the equivalent of the chatty computer on TV's Star Trek, the MXP 100 did an excellent job of pulling up our song requests. VoiceNav worked best when we spoke the entire file name, artist and song title, in the order it appeared on the screen.

If the unit had trouble discerning a phrase, it let out a short beep through the headphones as a request to repeat the phrase. We noticed it was easiest when you got a clear pronunciation on the first shot.

Under the best conditions, we would name a song in our normal speaking tone and pace and the unit would jump to the song. Sometimes when we called for that same song title we would get the beep instead. If we got a beep, the player required us to speak more deliberately and slowly otherwise it would beep again, suggesting VoiceNav drops down a gear, possibly to sample more intensely.

If the song has a one-word song title not repeated on another tune, the VoiceNav can jump to it just by saying that word. Because there is only one word to analyze, it works quicker too. Still, the unit was more accurate when you said the full track name. The more words you say, the more time the unit takes to digest and act on the command. Type A personalities take note.

It helps if the words in a title are actually in a dictionary. The track "Sum41 - Fat Lip" for example, "Sum41" is not a word. Surprisingly, when we said just the word "Sum41" the MXP 100 found the correct track almost every time. The same thing happened when we said just "U2" to find U2's One. Most impressive.

Testing VoiceNav -

Now realize that external ambient noise can affect results and so we tested VoiceNav under several conditions. We chose a mixture of song titles, some short and simple and others more apt to challenge VoiceNav like Sum41. Among the titles on our player as each track is written:

Detroit Cobras - I'll be home again
Howie Day - Babylon
Sam Philips - Love is everywhere you go
Wishing
Thee Midnighters - The town I live in
Sum41 - Fat lip
Sundays - Summertime
U2 - One
Sonics - Have love will travel
Robert Johnson - 32-20 Blues
Sebadoh - Willing to wait
Beach Boys - Kiss me baby
Lit - Miserable
Incubus - I wish you were here.

As soon as you enable VoiceNav the folder of where the tracks lie begins to blink. Saying "Track" causes the cursor to highlight and blink on the first track name. You say another track name to switch songs. A good way to judge how long it takes to switch is count how many times the cursor blinks on a song between when you call for the next track and when it actually appears.

Test 1 - Ideal conditions. Quiet room, MXP 100 is laying on the table. We speak directly to the player about a foot above the internal microphone. - Result: Superior if you say the entire song title, but order matters.

We were able to run through the entire list above twice - artist and song title said in the exact order they appear in the player - without a single miss. Average time it took to find and display the proper track was about 4 cursor blinks. The quickest were "Lit - Miserable" and "Wishing" at three blinks. The longest was "Robert Johnson - 32-20 Blues" at seven blinks. Finally, we called "Howie Day - Babylon" really quickly, practically mumbling the last word, before we threw the unit off. Excellent performance, better than we expected.

We then tried the same test just using either the song title or the artist alone and our results dropped to about 75%. Some artists had several songs on our player (Howie Day - Australia for example) posing a dilema for VoiceNav over which track to select when only the artists name was uttered. When VoiceNav did find the proper track successfully, it did so in about the same time as it took with the full title, an average of about 4 cursor blinks.

Order Matters

Now for a simple twist. We next spoke the full track title, but said the song title and artist name in reverse order as it appeared on the display. Here VoiceNav had trouble, giving us the beep on every single attempt. Order, it seems, matters with VoiceNav. We're guessing that it's because VoiceNav develops an expected vocal wave pattern, looking at the title as an entire phrase rather than word for word.

That could be a little frustrating to users who have tracks where the order of the artist and song name in the title is mixed. Still, when you get the order right the results are outstanding.

Test 2 - Walking outside with occasional traffic passing by. All track names said in proper order. - Result: very good to excellent.

When standing still, our results were nearly identical to those above. Saying the entire title name we got a hit almost every time. The only times we missed were when vehicles would pass as we spoke.

Next we repeated track names while walking. Here the results were very good, but the player sometimes needed an extra cursor blink or two to find the tracks. A couple of times it missed without traffic, but about 90-95% percent of the time it hit. A possible reason may be the motion from walking puts a slight quiver in our voices that make it a little harder for VoiceNav to understand. Still the unit performed excellently.

Because of the lessened performance we received under ideal conditions when we said only the artist or song name or when we said the track name in reverse order, we didn't bother to test that here. In those situations we already know the results will be hit or miss.

Test 3 - Difficult conditions. Using VoiceNav in a crowded gym. - Result: ambient noise proved too much for VoiceNav.

We tried and found that the background din of music, talking, and slamming weights was too much for VoiceNav. Once in a blue moon we got the track to shift, but not until speaking loud enough to draw the gaze of a few patrons who wondered why we were yelling at our MP3 player....

....The addition of VoiceNav is what truly differentiates the MXP 100 from the competition. VoiceNav proves to be a surprisingly excellent speech recognition system that works effectively at making the playing of your favorite songs more convenient. That makes all the difference between a useful tool and an expensive toy.

Speech recognition - as well as VoiceNav itself - has its limitations like any other technology, but with a little knowledge of how to get the most out of it, VoiceNav's performance was both impressive and compelling. I hardly have a radio voice, yet under the best of conditions the unit successfully pulled up every song I called out without fail (except the last when I intentionally garbled it). The system also performed great in real world conditions, providing there was not too much ambient noise.

We wish we had VoiceNav on our Archos Jukebox, a player we filled with tunes, but made the mistake of putting them all under one directory, making for a lot of scroll and search. I'll guess we will just have to take the time reload all the songs on that player, adding several folders to make searching easier.

http://www.mp3newswire.net/stories/2002/mxp100review.html