Voice annunciation using a publish/subscribe message broker
Original Publication Date: 2002-Oct-23
Included in the Prior Art Database: 2002-Oct-23
In many industrial monitoring and control applications, a well-known technique for alerting operators of exceptional alert conditions, is "voice annunciation". This is generally inflexible: the annunciation unit has to be physically close to the application which is generating the alert, and changing the "text" of an announcement typically involves changes to the application. The proposed annunciation system is based on a publish/subscribe message broker infrastructure, such as IBM's WebSphere MQ Integrator, and network-connected annunciation devices, which can be located anywhere there is network access. Specialised processing nodes in the broker perform text-to-speech processing, producing a set of "phoneme codes" from an input text string. Announcements are "published" by the alert application on a certain topic, which indicates the category of users who should hear the alert. The broker matches the topic with the subscriptions of the voice annunciation units, which have been configured to listen out for certain categories of announcements, and sends a copy of the phoneme string to each validated subscriber unit. Decoupling the application from the text to speech conversion enables many administrative changes to be applied to the whole system from a central location, removing the need for costly reprogramming.