Approaches to verifying the identity of speakers through objective, algorithmic characterizations of voice features and systems for implementing the same
Introduced here are approaches to dynamically verifying the identity of an individual and assessing the health of the individual through objective, algorithmic characterizations of voice features derived from audio recordings created by the individual. These voice features may be paralinguistic features, non-linguistic features, or linguistic features. By characterizing these voice features, insights can be gleaned into how events represented in the audio recordings should be emotionally characterized. For example, for each audio recording created by the individual, a health management platform may calculate a valence measure that is based on the voice features established for that audio recording. These valence measures can be documented, for example, in a profile maintained for the individual, such that changes in health can be detected in a more reliable, quantitative manner.
1. A method for assessing health of an individual using a distributed system that includes (i) a computer server on which a health management platform is executing and (ii) an electronic device on which a computer program is executing, the method comprising:
monitoring, by the computer program, a location of the electronic device so as to establish when the electronic device is located proximate to a predetermined location;
generating, by the computer program, a notification that prompts the individual to create an audio recording in response to a determination that the electronic device is located proximate to the predetermined location;
applying, by the computer program, a noise reduction algorithm to improve signal-to-noise (SNR) ratio of audio media that is recorded by the electronic device and that is representative of the audio recording created by the individual, prior to transmitting the audio media to the health management platform;
transmitting, by the computer program, the audio media to the health management platform over a network via which the electronic device is communicatively connected to the computer server;
processing, by the health management platform, the audio media using an algorithm that objectively characterizes a frequency-domain feature and produces, as output, a value for the frequency-domain feature;
verifying, by the health management platform, an identity of the individual based on a comparison of the value produced for the frequency-domain feature to a reference value that is known to be representative of the individual; and
in response to verifying the identity of the individual,
establishing whether the individual has experienced a change in health based on the value produced for the frequency-domain feature.
2. The method of claim 1 , wherein the frequency-domain feature is pitch, center formant frequency, shimmer, or jitter.
3. The method of claim 1 , further comprising:
parsing, by the health management platform, the audio media to identify portions, if any, that include a voice of someone other than the individual; and
removing, by the health management platform, the portions such that the algorithm is applied only to audio media that includes a voice of the individual.
4. The method of claim 1 ,
wherein the method further comprises:
applying, by the health management platform, a second algorithm that objectively characterizes a second frequency-domain feature and produces, as output, a second value for the second frequency-domain feature; and
wherein said determining and said establishing are further based on the second value produced for the second frequency-domain feature.
5. The method of claim 1 ,
wherein the method further comprises:
applying, by the health management platform, a second algorithm that objectively characterizes a time-domain feature and produces, as output, a second value for the time-domain feature; and
wherein said determining and said establishing are further based on the second value produced for the time-domain feature.
6. The method of claim 1 , further comprising:
storing, by the health management platform, the value produced for the frequency-domain feature in a data structure that includes prior values produced for the frequency-domain feature over time and that is representative of a digital profile maintained for the individual.
7. The method of claim 6 , wherein said establishing comprises determining that the value produced for the frequency-domain feature differs from at least one prior value produced for the frequency-domain feature by more than a predetermined amount, based on comparing the value produced for the frequency-domain feature to the prior values in the data structure.
8. The method of claim 1 , further comprising:
in response to a determination that the individual has experienced a change in health,
identifying, by the health management platform, interactive feedback that includes a recommendation for addressing the change in health; and
transmitting, by the health management platform, the interactive feedback to the electronic device for display to the individual by the computer program.
9. The method of claim 1 , further comprising:
in response to a determination that the individual has experienced a change in health,
transmitting, by the health management platform, an indication of whether the predetermined location provoked positive valence or negative valence to the electronic device for display to the individual by the computer program.