Method and device for instantly changing the speed of a speech

• US 6,205,420 B1
• Filed: 11/06/1998
• Issued: 03/20/2001
• Est. Priority Date: 03/14/1997
• Status: Expired due to Term

• Current Assignee: Nippon Hoso Kyokai
• Original Assignee: Nippon Hoso Kyokai
• Inventors: Imai, Atsushi, Ando, Akio, Takagi, Tohru, Seiyama, Nobumasa
• Primary Examiner(s): Dorvil, Richemond
• Assistant Examiner(s): Nolan, Daniel A.
• Application Number:
  US09/180,429
• Time in Patent Office: 865 Days
• Field of Search: 704/200, 704/211, 704/201, 704/267, 704/258, 704/271, 704/278, 704/266
• US Class Current: 704/211
• CPC Class Codes: G10L 21/04 Time compression or expansion

Abstract

An analysis processor applies an analysis process to input speech data thereby to obtain block lengths for respective attributes of voiced sound, voiceless sound and silence. A block data splitter splits the input speech data into blocks having the block lengths dependent on the respective attributes. A block data memory sequentially stores speech data split by the block data splitter as block speech data and the block lengths. A connection data generator generates connection data for connecting the adjacent block speech data each other at every moment by using the block speech data. A connection data storing portion sequentially stores the connection data. A connection order generator generates block connection order of the block speech data and the connection data at every moment according to at least the block lengths output sequentially from the block data storing portion and extension scaling factors in time for the respective attributes. A speech data connector connects sequentially the block speech data and the connection data based on the block connection order. Accordingly, the speed of output speech can be instantly changed in response to an instruction of an operator.

Claims

1. A method for instantly changing the speed of speech, comprising the steps of:

applying an analysis process to input speech data thereby to obtain block lengths for respective attributes of voiced sound, voiceless sound and silence;

splitting the input speech data having a voiced sound section, a voiceless sound section and a silent section into blocks having the block lengths dependent on the respective attributes;

storing the split speech data as block speech data and the block lengths sequentially in a buffer and outputting the block speech data and the block lengths sequentially from the buffer;

generating connection data at every moment, which are to be replaced or inserted between adjacent block speech data to connect the adjacent block speech data each other, every block, and then storing the connection data sequentially in another buffer and outputting the connection data sequentially from the other buffer;

generating block connection order of the block speech data and the connection data at every moment according to at least the block lengths output sequentially from the buffer and extension scaling factors in time for the respective attributes; and

connection sequentially the block speech data output from the buffer and the connection data output from the other buffer according to the block connection order to thus generate output speech data extended in time as compared with the input speech data.

2. A method for instantly changing the speed of speech according to claim 1, wherein the connection data are generated block by block by applying two windows to speech data located at a start portion of a concerned block and speech data located at a start portion of a succeeding block respectively, and then overlap-adding the start portion of the succeeding block to the start portion of the concerned block, each window having the shape of a predetermined line in a predetermined time interval.

3. A device for instantly changing the speed of speech, comprising:

an analysis processor for applying an analysis process to input speech data thereby to obtain block lengths for respective attributes of voiced sound, voiceless sound and silence;

a block data splitter for splitting the input speech data having a voiced sound section, a voiceless sound section and a silent section into blocks having the block lengths dependent on the respective attributes;

a block data storing portion for sequentially storing speech data split by the block data splitter as block speech data and the block lengths;

a connection data generator for generating connection data at every moment, which are able to be replaced or inserted between adjacent block speech data to connect the adjacent block speech data each other, by using the block speech data obtained by the block data splitter;

a connection data storing portion for sequentially storing the connection data being generated by the connection data generator;

a connection order generator for generating block connection order of the block speech data and the connection data at every moment according to at least the block lengths output sequentially from the block data storing portion and extension scaling factors in time for the respective attributes; and

a speech data connector for connecting sequentially the block speech data output from the block data storing portion and the connection data output from the connection data storing portion based on the block connection order obtained by the block connection order generator to thus generate output speech data extended in time as compared with the input speech data.

4. A device for instantly changing the speed of speech according to claim 3, wherein the connection data generator generates the connection data block by block by applying two windows to speech data located at a start portion of a concerned block and speech data located at a start portion of a succeeding block respectively, and then overlap-adding the start portion of the succeeding block to the start portion of the concerned block, each window having the shape of a predetermined line in a predetermined time interval.

5. A device for instantly changing the speed of speech according to claim 3, wherein the connection order generator includes,

a read/write memory for storing the extension scaling factors in time for the respective attributes, and

a connection order deciding processor for reading the the extension scaling factors in time for the respective attributes stored in the read/write memory at a predetermined time interval, and generating the block connection order of the block speech data and the connection data at every moment based on the extension scaling factors, the block lengths output from the block data storing portion, and the already-connected information output from the speech data connector.