NJ Transit receives its train and bus data from the IPS rail database management system, which provides feedback from track circuits and other real-time information including train locations. That information is fed into NJ Transit’s various server/database platforms on a continuous basis, with hundreds of new updates per hour. NJ Transit operators then select the information to broadcast over public address and signage.
The new design calls for NJ Transit to soon use a single system for all aspects of its intermodal transportation systems that is easier to maintain, operate, amend and update.
“As time goes on we’ll have more and more control over the different aspects, and fewer contracts with outside vendors,” said William Loucas, Technical Specialist, Public Address Communications Technologies for NJ Transit. “For now, we have standardized the doorway that all of these systems come into, which means that as one vendor leaves, it is quick and easy to eliminate that part of the system.”
The flow of passenger information begins to take shape after NJ Transit operators mine the appropriate data from the IPS system and enter it into the NJ Transit database system. The messages appearing on signs at the various stations and the NJ Transit Web site are converted to voice using an automated text-to-speech engine that produces a robotic voice message. While this may bring visions of an old Kraftwerk album, the result is extraordinarily lifelike and natural. Delivery of the text and audio messages to each station continues throughout the day on a scheduled basis.
NJ Transit selected the Loquendo Text-To-Speech (TTS) software engine to convert the database information for public address. Rountree notes that many text-to-speech engines are very cumbersome and challenging to fine-tune. Many also burden the user with reams of programming tools that are required to even slightly tweak the system.
“Loquendo has a very simplistic approach to adjust cadences and punctuations,” he said. “It’s fairly easy to alter the operation of the engine and how it relates to speech. It’s faster to correct, faster to set up and it supports 16-bit resolution audio streams to more effectively reproduce the original message as voice.”
Many text-to-speech systems are designed for telephone and consequently work fine for audio conversions using 8-bit resolution. At twice the bit resolution, the Loquendo system essentially provides far more information in the same slice of bandwidth to reproduce voice at far greater quality.
“8-bit resolution is simply insufficient for full-range audio at a rail or bus station,” added Rountree. “Our peak intelligibility is greatly enhanced with 16-bit resolution streams. The quality is much better because we can adjust the amount of data we are putting into the voice we are streaming.”
The Audio Element
The IP network infrastructure that starts with the database systems and the Loquendo engine extends onward to the audio distribution, processing and playout at each location.
The text-to-speech process produces a WAV file that is converted to mp3 for Web distribution en route to a Barix Exstreamer 100 IP decoder. The Exstreamers are IP-addressable, and mp3 files are sent to one or more devices assigned to each specific message.
Generally speaking, there is one Exstreamer at each location to receive the streams. Each decoded stream is passed to the local audio processing system before playing out over the station PA system. (The text version of the same message is simultaneously displayed on nearby signs, making it ADA-compliant; and also posted to the NJ Transit Web site.)
Rountree first discovered Barix through his work with Ramsay, which dates back nearly a decade when the two collaborated on speech generation system for a new rail transfer station in Secaucus, N.J., just south of New York City. Ramsay’s experience with Barix also dates back several years, including Web jukebox applications where his clients would use Barix encoding devices to stream music over the Web.
“Barix gave NJ Transit both a very good price point for audio distribution as well an openly programmable box that can use different firmware to build custom functionality,” said Ramsay. “They can point this box in different directions for different operations. We considered other IP delivery systems but they would have been more expensive and more difficult to customize and change.”