System and method for secure processing of speech signals using pseudo-speech representations
A method, computer program product, and computing system for processing a speech signal. A sensitive portion of the speech signal is identified. A pseudo-speech representation of the sensitive portion is generated using a voice converter system. Speech processing is performed on the speech signal and the pseudo-speech representation of the sensitive portion using a speech processing system.
1 . A computer-implemented method comprising:
receiving, by a speech privacy processor, a speech signal recorded during a medical encounter between a medical professional and a patient, wherein a sensitive portion of the speech signal includes Personally Identifiable Information (PII) of the patient;
processing, by the speech privacy processor, the speech signal to obtain a synthetic speech signal at least by converting the PII within the sensitive portion of the speech signal into a pseudo-speech representation of the PII in the synthetic speech signal using a voice converter, wherein the voice converter reduces an acoustic quality of one or more linguistic units associated with the PII by filtering the sensitive portion of the speech signal in a modulation domain to remove modulations above a predefined threshold, thereby generating the pseudo-speech representation of the PII, wherein the modulation domain includes carrier signals and modulators; and
transmitting, by the speech privacy processor, the synthetic speech signal to an automatic speech recognition (ASR) processor.
2 . The computer-implemented method of claim 1 , wherein the PII within the sensitive portion of the speech signal is converted into the pseudo-speech representation based on a predefined acoustic signature.
3 . The computer-implemented method of claim 2 , wherein the predefined acoustic signature is based upon, at least in part, a content type of the sensitive portion of the speech signal.
4 . The computer-implemented method of claim 1 , wherein converting the PII within the sensitive portion of the speech signal into the pseudo-speech representation includes transforming an articulatory feature of the sensitive portion of the speech signal.
5 . The computer-implemented method of claim 1 , wherein converting the PII within the sensitive portion of the speech signal into the pseudo-speech representation includes performing a spectral modification of the sensitive portion of the speech signal.
6 . The computer-implemented method of claim 1 , further comprising:
marking the pseudo-speech representation with a watermark.
7 . A computing system comprising:
a processor; and
a memory storing programming instructions for execution by the processor, the programming instructions, upon execution by the processor, causing the computing system to perform the following steps:
receiving, by a speech privacy processor, a speech signal recorded during a medical encounter between a medical professional and a patient, wherein a sensitive portion of the speech signal includes Personally Identifiable Information (PII) of the patient;
processing, by the speech privacy processor, the speech signal to obtain a synthetic speech signal at least by converting the PII within the sensitive portion of the speech signal into a pseudo-speech representation of the PII in the synthetic speech signal using a voice converter, wherein the voice converter reduces an acoustic quality of one or more linguistic units associated with the PII by filtering the sensitive portion of the speech signal in a modulation domain to remove modulations above a predefined threshold, thereby generating the pseudo-speech representation of the PII, wherein the modulation domain includes carrier signals and modulators; and
transmitting, by the speech privacy processor, the synthetic speech signal to an automatic speech recognition (ASR) processor.
8 . The computing system of claim 7 , wherein the PII within the sensitive portion of the speech signal is converted into the pseudo-speech representation based on a predefined acoustic signature.
9 . The computing system of claim 8 , wherein the predefined acoustic signature is based upon, at least in part, a content type of the sensitive portion of the speech signal.
10 . The computing system of claim 7 , wherein converting the PII within the sensitive portion of the speech signal into the pseudo-speech representation includes transforming an articulatory feature of the sensitive portion of the speech signal.
11 . The computing system of claim 7 , wherein converting the PII within the sensitive portion of the speech signal into the pseudo-speech representation includes performing a spectral modification of the sensitive portion of the speech signal.
12 . The computer-implemented method of claim 1 , wherein the ASR processor is located remotely from the speech privacy processor, and wherein the synthetic speech signal is transmitted over a network that communicatively couples the speech privacy processor to the ASR processor.
13 . The computer-implemented method of claim 1 , wherein reducing the acoustic quality of the one or more linguistic units associated with the PII renders the pseudo-speech representation of the PII generally unintelligible during speech recognition by the ASR processor.
14 . The computer-implemented method of claim 1 , wherein the voice converter reduces the acoustic quality of the one or more linguistic units associated with the PII by shortening a duration of the one or more linguistic units associated with the PII.
15 . The computer-implemented method of claim 1 , wherein the voice converter reduces the acoustic quality of the one or more linguistic units associated with the PII by reducing a clarity of articulation of the one or more linguistic units associated with the PII.
16 . A computer program product residing on a non-transitory computer readable medium having programming instructions stored thereon which, when executed by a processor of a system, cause the system to perform the following operations:
receiving, by a speech privacy processor, a speech signal recorded during a medical encounter between a medical professional and a patient, wherein a sensitive portion of the speech signal includes Personally Identifiable Information (PII) of the patient;
processing, by the speech privacy processor, the to obtain a synthetic speech signal at least by converting the PII within the sensitive portion of the speech signal into a pseudo-speech representation of the PII in the synthetic speech signal using a voice converter, wherein the voice converter reduces an acoustic quality of one or more linguistic units associated with the PII by filtering the sensitive portion of the speech signal in a modulation domain to remove modulations above a predefined threshold, thereby generating the pseudo-speech representation of the PII, wherein the modulation domain includes carrier signals and modulators; and
transmitting, by the speech privacy processor, the synthetic speech signal to an automatic speech recognition (ASR) processor.
17 . The computer program product of claim 16 , wherein the voice converter reduces the acoustic quality of the one or more linguistic units associated with the PII by shortening a duration of the one or more linguistic units associated with the PII.
18 . The computer program product of claim 16 , wherein the voice converter reduces the acoustic quality of the one or more linguistic units associated with the PII by reducing a clarity of articulation of the one or more linguistic units associated with the PII.