Skip Navigation Link www.weather.gov
NOAA logo - Click to go to the NOAA homepage National Weather Service NWS logo - Click to go to the NWS homepage
NOAA Weather Radio All Hazards
 
   

Voice Improvement Processor

NWR Voice Improvement Status

The National Weather Service implemented new, improved voices for NOAA Weather Radio (NWR). These voices are more understandable and human-sounding than the previous voice. New male and female voices were implemented nationwide on NWR at NWS Forecast Offices in 2002.

History

The current PC based broadcast system was deployed in January 1997. The system offered an enormous advance in speed in delivering warnings, watches, forecasts and other hazard information 24 hours a day. Before that time, forecasters manually recorded and transmitted each of message. The computerized system enables the National Weather Service to send out multiple warnings over multiple transmitters simultaneously up to several minutes faster. These minutes save lives.

While automating NWR was a great advance in public safety, there were complaints that the voice was hard to understand. In August 2000, John J. Kelly, Director of the National Weather Service, set a goal to improve voice quality on NWR.

A voice improvement team was formed, and a Request For Information went out to the industry to explore what Text to Speech technology was available. Input was received from a broad range of industry representatives, and from the Information Technology Laboratory in the National Institute of Science and Technology (NIST). The National Weather Service held meetings with industry to discuss available technology and NOAA Weather Radio requirements.

One option studied was use of prerecorded words and phrases for all watches, warnings, and advisories. This method works for highly standardized and limited vocabulary. It was decided that this type of system could not handle the wide range of hazard and geographical information provided by the National Weather Service in the NOAA Weather Radio broadcasts.

It was agreed that NOAA Weather Radio would require a sophisticated Text to Speech software with virtually unlimited vocabulary to provide the full range of watches, warnings, advisories, forecasts, synopses and other hazard information necessary to public safety.

A Request for Proposals -- 52-DDNW-1-90021 -- was published for bids on a sophisticated Text to Speech software. The National Weather Service evaluated the proposals received on the basis of voice quality, speed of system, integration of voice with existing console system, and cost.

After months of evaluating voice technologies and receiving public input, including over 19,000 Internet survey comments, the National Oceanic and Atmospheric Administration (NOAA) awarded Siemens Information and Communication Network of Boca Raton, Florida, a contract for the voice improvement software product, known as Speechify in August 2001. Siemens has teamed with SpeechWorks International of Boston, Mass., to provide software that combines phonetic sounds with natural language modeling. This software combines concatenated, pre-recorded phonetic sounds with the emphasis and intonation of a human voice. Speechify voices received the most favorable comments in the web page public opinion survey.

The Speechify commercial product was integrated with supporting hardware and software. Testing was conducted with the existing automated broadcast console at NWS Headquarters and at five field offices across the country. The new NOAA Weather Radio voices debuted on the air nationwide in May and June 2002. Listeners will ultimately hear current weather conditions and severe weather warnings delivered in the new computer-generated voices. Word and geographic name pronunciations can be modified at each local NWS Weather Forecast Office to further improve understandability.



NOAA, National Weather Service
Office of Climate, Water, and Weather Services
1325 East West Highway
Silver Spring, MD 20910
Questions, Comments?

Disclaimer Privacy Policy

 

`