Compact, verifiable record of an audio communication and method for making same
A compact, self-authenticating, and speaker-verifiable record of an audio communication involving one or more persons comprises a record, encoded on a non-transitory, computer-readable medium, that consists essentially of: a voiceprint for each person whose voice is encoded in the record; a plurality of transcription records, where each transcription record consists essentially of a computer-generated speech-to-text decoding of an utterance and voiceprint associating information that associates a speaker of the utterance with one of the voiceprints stored in the record; and self-authenticating information sufficient to determine whether any of the information encoded in the communication record has been altered.
1. A compact, self-authenticating, and speaker-verifiable record of an audio communication involving one or more persons, said communication record encoded on a non-transitory, computer-readable medium and consisting essentially of:
a voiceprint for each person whose voice is encoded in said record;
a plurality of transcription records, each transcription record consisting essentially of a computer-generated speech-to-text (STT) decoding of an utterance and voiceprint associating information that associates a speaker of said utterance with one of the voiceprints stored in said record; and,
self-authenticating information sufficient to determine whether any of the information encoded in said communication record has been altered.
2. A communication record, encoded on a non-transitory, computer-readable medium, as defined in claim 1 , wherein said communication record further consists essentially of:
meta-data associated with each transcription record.
3. A communication record, encoded on a non-transitory, computer-readable medium, as defined in claim 2 , wherein said meta-data includes a confidence score that indicates a likelihood that the corresponding transcription record contains an accurate STT decoding of the corresponding utterance.
4. A communication record, encoded on a non-transitory, computer-readable medium, as defined in claim 2 , wherein said meta-data further includes at least one of:
a timestamp that indicates a starting time of the corresponding utterance;
a duration of the utterance;
role identification information that indicates the status of the utterance speaker as an agent or a caller;
utterance overtalk information;
utterance volume information;
acoustically derived utterance emotion information;
automatic number identification (ANI) information; and,
dialed number identification system (DNIS) information.
5. A communication record, encoded on a non-transitory, computer-readable medium, as defined in claim 1 , wherein said self-authenticating information comprises a message authentication code (MAC).
6. A communication record, encoded on a non-transitory, computer-readable medium, as defined in claim 5 , wherein said MAC comprises one of an MD5, SH-1, SH-2, or SH-3 code computed from the contents of said communication record.
7. A communication record, encoded on a non-transitory, computer-readable medium, as defined in claim 1 , wherein said communication record includes at least two voiceprints.
8. A communication record, encoded on a non-transitory, computer-readable medium, as defined in claim 7 , wherein said voiceprints are derived using one of: frequency estimation, hidden Markov models, Gaussian mixture models, pattern matching algorithms, neural networks, matrix representation, vector quantization, or decision trees.
9. A communication record, encoded on a non-transitory, computer-readable medium, as defined in claim 7 , wherein said communication record further consists essentially of audio data from which each voiceprint in said communication record was derived.
10. A communication record, encoded on a non-transitory, computer-readable medium, as defined in claim 3 , wherein said communication record further consists essentially of audio data whose corresponding transcription confidence score is below a predetermined likelihood of accuracy threshold.
11. A process for creating a compact, self-authenticating, and speaker-verifiable record of an audio communication session involving one or more persons, said process comprising:
creating a voiceprint for each new person whose voice is detected during the communication session;
performing automatic speech recognition (ASR) decoding to create a plurality of transcription records, each transcription record consisting essentially of an ASR decoded utterance and voiceprint associating information that associates a speaker of said utterance with one of the previously created voiceprints; and,
creating the compact, self-authenticating, and speaker-verifiable record of the audio communication session by storing the voiceprints and transcription records, along self-authenticating information sufficient to determine whether any of them have been altered, in a non-transitory, computer-readable medium.
12. A process, as defined in claim 11 , further comprising: associating meta-data with each transcription record; and,
storing said meta-data, along with the voiceprints, transcription records, and self-authenticating information, in the non-transitory, computer-readable medium.
13. A process, as defined in claim 12 , wherein:
associating meta-data with each transcription record comprises computing and associating a confidence score that indicates a likelihood that the corresponding transcription record contains an accurate ASR decoding of the corresponding utterance.
14. A process, as defined in claim 12 , wherein:
associating meta-data with each transcription record comprises associating at least one of:
a timestamp that indicates a starting time of the corresponding utterance;
a duration of the utterance;
role identification information that indicates the status of the utterance speaker as an agent or a caller;
utterance overtalk information;
utterance volume information;
acoustically derived utterance emotion information;
automatic number identification (ANI) information; and,
dialed number identification system (DNIS) information.
15. A process, as defined in claim 11 , wherein said self-authenticating information comprises a message authentication code (MAC).
16. A process, as defined in claim 15 , wherein said MAC comprises one of an MD5, SH-1, SH-2, or SH-3 code computed from the contents of said record.
17. A process, as defined in claim 11 , wherein the process creates at least two voiceprints.
18. A process, as defined in claim 17 , wherein creating the voiceprints involves use of: frequency estimation, hidden Markov models, Gaussian mixture models, pattern matching algorithms, neural networks, matrix representation, vector quantization, or decision trees.
19. A process, as defined in claim 17 , wherein:
creating the compact, self-authenticating, and speaker-verifiable record of the audio communication session involves storing the voiceprints, transcription records, and audio data from which each voiceprint was derived, along self-authenticating information sufficient to determine whether any of them have been altered, in the non-transitory, computer-readable medium.
20. A process, as defined in claim 13 , wherein for any audio data whose corresponding transcription confidence score is below a predetermined likelihood of accuracy threshold, said process further involves storing said audio data, along with said meta-data, voiceprints, transcription records, and self-authenticating information, in the non-transitory, computer-readable medium.