Speech recognition will be used to add command and control features that help consumers navigate through the thousands of songs that can be stored on digital audio players equipped with hard-disk drives, suppliers said. A speech engine would recognize a set of prefabricated commands that control the searching, playing and listing of stored songs.

Portable MP3 players that rely on flash memory to store a small number of songs accounted for the lion's share of digital audio player shipments in 2000, but players with hard-disk drives are a growing segment. Several manufacturers, including Hango Electronics of South Korea, have introduced or announced plans for drive-equipped audio players and jukeboxes that store tens of gigabytes worth of music, far more than flash-based players.

Some of these companies are talking with Sensory about plans to add speech recognition to their audio jukeboxes and portables, said Erik Soule, director of marketing for Sensory (Santa Clara, Calif.), which sells speech recognition chips and software for toys and consumer electronics.

Deals with OEMs are "weeks, not months" from release, said Soule, who declined to name companies that have signed contracts for Sensory's technology.

Those contracts could involve Sensory's Voice Activation software, designed to add speaker-dependent or speaker-independent recognition capabilities, speech prompting, and verification to embedded consumer electronics systems. Sensory also offers speech recognition microcontrollers for use in toys and other goods.

Logical step

Speech recognition is "a logical way to navigate through the anywhere from 2,000 to 4,000 songs" stored on a hard-disk, said Fred Falk, president and chief executive of e.Digital Corp. (San Diego), a developer of DSP-based digital audio player reference designs that it licenses to manufacturers.

E.Digital will embed speech recognition into its next-generation hard-disk-based digital audio player reference platform, scheduled for introduction in the third or fourth quarter, Falk said. The company licenses its speech-recognition technology from a company it wouldn't identify.

The reference design will implement non-user-dependent speech recognition, and will allow any user to push a button and say a word in a song, name an artist or album title, prompting a search through the song database and then to list or play the first song it finds.

That capability uses software that's been designed for desktop or laptop computers and ported to a DSP in the portable player design, Falk said.

The e.Digital reference platform includes a digital signal processor from Texas Instruments Inc. (Dallas) and supports several digital audio formats, including MP3, AAC, Microsoft's Windows Media Audio and Q-Design's QDX codec.

The speech-enabled audio player will be e.Digital's second audio platform with disk-based storage. The company completed its first design last year and licensed it to several consumer electronics OEMs, including Hango of South Korea and Eastec of Taiwan.

Request Multimedia Inc. (Troy, N.Y.), which sells a home stereo MP3 jukebox over the Web, has a speech-enabled jukebox player on its product road map, but its introduction may be a year or so away.

"We are considering a number of different speech technologies," said Steven Vasquez, Request Multimedia's chief executive. "Some are ready for prime time, and some are not."

In addition to speech recognition features, Request is optimizing navigation for music searches and adding connectivity for certain portables. The AudioReQuest-II MP3 unit, which stores over 10,000 songs on a 30-Gbyte Quantum hard drive, can load music from CDs and PCs, and record from audio cassettes, LPs and radio. The player includes Ethernet, USB, and parallel connectivity ports, and will be available this fall for $1,600, the company said.