Delay computation for real-time synchronization of speech and its converted text

Transmission of real-time text data integrated with other multimedia applications such as audio and video has raised the issues of compatibility and synchronization among these applications since a stringent quality of service (QoS) guarantee is especially critical for real-time traffic. In order to meet the real-time properties, text must be produced efficiently to integrate its transmission with other multimedia applications. Literature survey shows that the text for the real-time transmission can be produced by different input sources such as from handwriting recognition, voice recognition or it can be entered by human users from a keyboard or any other input method [1]. This paper presents an efficient way of producing text which to the best of our knowledge has not been previously explored. We propose to generate text from the recognition of the real-time voice in the source machine. We calculate the delay in the speech-recognition or speech-to-text conversion. Based on these statistics we suggest a buffer size to store the voice data until its respective text is generated. This enables us to transmit both voice and its converted text synchronously. We find, and show it graphically, that this delay is almost negligible and there is almost no queue formation in the buffer. Hence both of the applications can be transmitted instantly that is as they are available. This research is a reasonable advancement in the subject area.