Browse Prior Art Database

Method for Identifying and Coordinating Digitally Recorded Audio Texts

IP.com Disclosure Number: IPCOM000110020D
Original Publication Date: 1992-Oct-01
Included in the Prior Art Database: 2005-Mar-25
Document File: 4 page(s) / 165K

Publishing Venue

IBM

Related People

Irvin, DR: AUTHOR

Abstract

The technique described here enables a machine to identify a digitally-recorded audio text from its contents in the absence of any other identifying information and to associate with that text a pre-established command list selected from a set of such lists stored efficiently in a random-access memory (RAM). The association is made by hashing 32,000-byte blocks of the audio text by means of a 32-bit cyclic redundancy check (CRC) in order to map the audio text to a shorter record (here, the frame-check sequence (FCS) that results from the CRC operation) that serves as a tag.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 44% of the total text.

Method for Identifying and Coordinating Digitally Recorded Audio Texts

       The technique described here enables a machine to
identify a digitally-recorded audio text from its contents in the
absence of any other identifying information and to associate with
that text a pre-established command list selected from a set of such
lists stored efficiently in a random-access memory (RAM).  The
association is made by hashing 32,000-byte blocks of the audio text
by means of a 32-bit cyclic redundancy check (CRC) in order to map
the audio text to a shorter record (here, the frame-check sequence
(FCS) that results from the CRC operation) that serves as a tag.  The
resulting tag is used to associate the audio text with a command list
that contains instructions or a pointer to instructions that may be
used to indicate how the audio text is to be decoded or how it is
to be associated with other stored documents such as
digitally-encoded alphabetic text or digitally-encoded images. In the
interest of brevity, all of these options are henceforth called
"command lists."

      With the help of the technique described here, the user creates
a Table held in RAM that contains the tags and command lists needed
to manage his particular library of audio texts.  Fig. 1 shows the
structure of this Table; the rows of the Table correspond to vectors
of data.  The first field of the row or vector contains a tag
generated by hashing.  The second field contains a two-bit codeword
that indicates the status of the tag that appears in the first field.
Codeword '00' indicates an unused or idle tag; codeword '01'
indicates a valid tag; codeword '10' indicates an invalid tag;
codeword '11' is not used. The third field contains the command list
associated with the tag that appears in the first field.

      Fig. 2 shows how new entries are loaded into the Table.  The
process begins when the operator enters the recorded audio text into
its reader and enters (by means of a keyboard or by means of
automated data transfer) the command list he wishes to associate with
this audio text into the top pf a FIFO (first-in, first-out) stack.
Fig. 2 shows that the progress of the algorithm is gated by the
judgement of an activity detector.  This detector is included in
order to accommodate a wide class of applications for this technique
wherein the audio text represents digitally-encoded speech or music.
In these cases, the initial blocks of the audio text may be recorded
noise with little information content.  The function of the activity
detector is to enable the larger algorithm in which it operates to
skip over these periods of noise in order to locate the beginning of
the useful message and thereby increase the probability of generating
an acceptable tag for this text.  One way in which such an activity
detector may be constructed is described in (1).

      The tag used to identify a particular audio text is created by
hashing sequential blocks of the acti...