Reducing perceived effects of non-voice data in digital speech
AssigneeDigital Voice Systems, Inc.
InventorJohn C. Hardwick
Life of this patent
Estimating the term…
Abstract
Non-voice data is embedded in a voice bit stream that includes frames of voice bits by selecting a frame of voice bits to carry the non-voice data, placing non-voice identifier bits in a first portion of the voice bits in the selected frame, and placing the non-voice data in a second portion of the voice bits in the selected frame. The non-voice identifier bits are employed to reduce a perceived effect of the non-voice data on audible speech produced from the voice bit stream.
Claims20
A method for embedding non-voice data into a voice bit stream, said bit stream comprised of one or more frames of voice bits, the method comprising the steps of:
- selecting a frames of voice bits to carry the non-voice data;
- and
- replacing at least a portion of the voice bits in the selected frame with non-voice data,
- wherein said replacement causes the selected frame to be identified as an invalid voice frame, and wherein quantizer state information associated with the voice bit stream is not updated for the selected frame.
A method for recovering non-voice data from a voice bit stream, the bit stream including one or more frames of voice bits, the method comprising:
- comparing the voice bits in a first portion of a frame against one or more predetermined patterns of non-voice identifier bits;
- and
- recovering non-voice data from a second portion of the frame when one of said predetermined patterns of non-voice identifier bits is detected;
- wherein said predetermined patterns of non-voice identifier bits reduce the perceived effect of the non-voice data on audible speech produced using the voice bit stream.
A method for recovering non-voice data from a voice bit stream used to produce a sequence of voice parameters used to generate audible speech, the bit stream including one or more frames of bits, the method comprising:
- a) decoding a frame of bits to produce voice parameters using a voice decoding method that employs quantizer state information;
- b) determining validity of the frame from the voice parameters;
- and
- c) upon determining that the frame is invalid, recovering non-voice data from the frame, repeating voice parameters decoded from a prior frame in the sequence of voice parameters, and not updating the quantizer state information for the frame.
A speech decoder configured to recover non-voice data from a voice bit stream, the bit stream including one or more frames of voice bits, the speech decoder being configured to:
- compare the voice bits in a first portion of a frame against one or more predetermined patterns of non-voice identifier bits;
- and
- recover non-voice data from a second portion of the frame when one of said predetermined patterns of non-voice identifier bits is detected;
- wherein said predetermined patterns of non-voice identifier bits reduce the perceived effect of the non-voice data on audible speech produced based on the voice bit stream.
Dependencies are inferred from the claim text.