Browse Prior Art Database

Method of Starting and Monitoring Tasks in a Multitasking Speech Recognition System

IP.com Disclosure Number: IPCOM000110235D
Original Publication Date: 1992-Nov-01
Included in the Prior Art Database: 2005-Mar-25
Document File: 2 page(s) / 80K

Publishing Venue

IBM

Related People

De Gennaro, S: AUTHOR [+2]

Abstract

There are many challenges to use multi-tasking for speech recognition. One such challenge is starting all the tasks, setting up all the interprocess communications (IPC), and monitoring the tasks so that should one terminate, it can either be restarted or the other tasks can be terminated, in order to provide continuity of servers or graceful shutdown. This article discloses using a program to establish IPC, spawning the individual tasks to perform the work and monitoring the tasks for abnormal completion.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 52% of the total text.

Method of Starting and Monitoring Tasks in a Multitasking Speech Recognition System

       There are many challenges to use multi-tasking for speech
recognition.  One such challenge is starting all the tasks, setting
up all the interprocess communications (IPC), and monitoring the
tasks so that should one terminate, it can either be restarted or the
other tasks can be terminated, in order to provide continuity of
servers or graceful shutdown.  This article discloses using a program
to establish IPC, spawning the individual tasks to perform the  work
and monitoring the tasks for abnormal completion.

      In the Tangora Automatic Speech Recognition System, there are
many separate processes (currently 8), which communicate
asynchronously with each other through message queues.  In addition,
there could be more than one speech recognizer running on a
particular machine.  Thus, a dynamic method of starting and
monitoring all the processes is required.  This article discloses
using a monitoring program, which spawns child  processes,
establishes IPC, and monitors its children for termination in the
Tangora Automatic Speech Recognizer (ASR).  This program is named
mama, since it spawns child processes.

      Thus, mama is a program which starts and monitors the Tangora
ASR, which has the following features:
o    Uses a process file that indicates the programs to be started.
o    Allows input message queues, output message queues, and shared
memory segments to be optionally created for each program to be
started.
o    Allows message queues to be shared among different programs.
o    Allows specifying the shared memory segment size for a program
that wants shared memory.
o    Allows specifying of w...