Browse Prior Art Database

A method to recover audio packet lost due to insufficient bandwidth by converting audio to text

IP.com Disclosure Number: IPCOM000247305D
Publication Date: 2016-Aug-19
Document File: 9 page(s) / 509K

Publishing Venue

The IP.com Prior Art Database

Abstract

Our invention is a method to recover audio packet loss by converting audio to text. This method will adopt speech to text technology to convert into text & send. Then speak out the text at the receiver side. It is a solution to transfer audio packets over a bad connection (low bandwidth or unstable connection) without disrupting the VoIP communication. Every word from the speaker is ensured to deliver to the receiver side even if there are packets lost during the transfer.

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

Page 01 of 9

A method to recover audio packet lost due to insufficient bandwidth by converting audio to text

Nowadays, VoIP (Voice over IP) communication is very common, especially among mobile device users. However the signal of wireless connection may not always be strong and stable, furthermore the bandwidth of the wireless connection may not always be big enough for VoIP communication. Packets may be lost with a bad connection, which causes the VoIP communication to be disrupted. In this invention, we provide a method to transfer audio packets over a bad connection without disrupting the VoIP communication.

We proposed a method to recover audio packet loss by converting audio to text. This method will adopt speech to text technologyto reduce the storage size, and then speak out the text at the receiver side.

Claim point:


1. Recover audio packet loss with lower storage size by converting speech to text, and store into next packet.

2. To recover the lost audio packet, the receiver will find the next audio packet, and extract the recovering text from the next packet, then speak out the text via machine speech technique.

How does it work:

When a speaker initiates an audio transmission, his/her voice is stored in both audio form and text format.

If the transmission becomes unstable and the bandwidth becomes limited (which ends up in packet loss), the text format stored in next package will be sent to the receiver side instead. The machine will read out the text form of the audio using machine speech technique.

Implementation:

Our invention will combine multiple packets into a segment to form a word.

In each segment, there is an audio form of the speaker's voice and a text form of the speaker's previoussegment voice.

If in any case one of the segment has significant amount of packet lost, then the text form of the current voice stored in nextsegment will be processed and speak out using machine speech on the receiver side.

Since all voice in every segment will have a copy of their text form stored on the speaker side, if more than one segments are lost, the successor segment will play out all previous packet's text form using machine speech.

For how to determine the speaker finishes a word:

If our invention determines the speaker finishes a word, it will convert that word(current segment) into...