Desired signal spotting in noisy, flawed environments
There are disclosed devices, system and methods for desired signal spotting in noisy, flawed environments by identifying a signal to be spotted, identifying a target confidence level, and then passing a pool of cabined arrays through a comparator to detect the identified signal, wherein the cabined arrays are derived from respective distinct environments. The arrays may include plural converted samples, each converted sample include a product of a conversion of a respective original sample, the conversion including filtering noise and transforming the original sample from a first form to a second form. Detecting may include measuring a confidence of the presence of the identified signal in each of plural converted samples using correlation of the identified signal to bodies of known matching samples. If the confidence for a given converted sample satisfies the target confidence level, the given sample is flagged.
1. A method of desired signal spotting in noisy, flawed environments comprising:
identifying a signal to be spotted;
identifying a target confidence level;
passing a pool of cabined arrays through a comparator to spot the identified signal, wherein the cabined arrays are derived from respective distinct environments, the cabined arrays comprising plural converted samples, and wherein spotting the identified signal comprises measuring a confidence of a presence of the identified signal in each of plural converted samples using correlation of the identified signal to bodies of known matching samples;
if the confidence for a given converted sample satisfies the target confidence level, flagging the given converted sample; and
the comparator re-correcting mislabeled manual inputs by automatically inputting text corrected per the comparator as the flagging for the given converted sample, wherein the mislabeled manual inputs are feeding back mislabeled corrections of mis-flagging from a human;
wherein each given converted sample comprises a product of a conversion of a respective original sample, the conversion including filtering noise and transforming the original sample from a first form to a second form.
2. The method of claim 1 wherein re-correcting includes performing a weighting analysis that determines that the mislabeled manual inputs are mislabeled or suspicious based on call information of the mislabeled manual inputs.
3. The method of claim 1 wherein the bodies of known matching samples are derived from a combination of manually selected samples and samples selected by the comparator.
4. The method of claim 1 wherein the conversion further includes anonymization.
5. The method of claim 1 wherein:
the respective distinctive environments comprise electronic waveforms of the respective original samples;
the noisy, flawed environments comprise at least one of noise in a call signal of the pool of cabined arrays or an erroneous sample of the bodies of known matching samples;
the signal to be spotted comprises one or more spoken terms; and
the first form is an audio electronic waveform and the second form is text.
6. The method of claim 1 further comprising:
pooling plural cabined arrays derived from the distinct environments with the identified signal to be spotted, the cabined arrays comprising plural call samples; and
converting each respective original sample to a converted sample.
7. An apparatus for desired signal spotting in noisy, flawed environments comprising:
a first identifier unit to identify a signal to be spotted;
a second identifier unit to identify a target confidence level;
a synchronizer coupled to the first and second identifier units to synchronize a pool of cabined arrays of call samples with the identified signal and the identified confidence level; and
a tester coupled to the synchronizer to pass the pool of cabined arrays of calls samples through a comparator;
the comparator to spot the identified signal, wherein the cabined arrays are derived from respective distinct environments, the cabined arrays comprising plural converted samples, and wherein spotting the identified signal comprises measuring a confidence of a presence of the identified signal in each of plural converted samples using correlation of the identified signal to bodies of known matching samples;
a flagger to flag a given converted sample if the confidence for the given converted sample satisfies the target confidence level;
a feedback loop to feed back into the comparator manual inputs that are mislabeled corrections of mis-flagging using inputted text as the flagging for the given converted sample; and
a corrector to re-correct the mislabeled corrections by automatically inputting text as the flagging for the given converted sample, wherein the mislabeled corrections are mislabeled correction of mis-flagging from a human;
wherein each given converted sample comprises a product of a conversion of a respective original sample, the conversion including filtering noise and transforming the original sample from a first form to a second form.
8. The apparatus of claim 7 wherein re-correcting includes performing a weighting analysis that determines that the mislabeled corrections are mislabeled or suspicious based on call information of the mislabeled corrections.
9. The apparatus of claim 7 further comprising a deriver to derive the bodies of known matching samples from a combination of manually selected samples and samples selected by the comparator.
10. The apparatus of claim 7 further comprising an anonymizer to anonymize the converted sample.
11. The apparatus of claim 7 wherein the first and second identifier units further comprise a user input device; and
wherein the comparator further comprises a predictive model for performing a weighting analysis comparison of the identified signal to bodies of known matching samples.
12. A comparator to spot desired signals in noisy, flawed environments comprising:
a measurer to measure a confidence of a presence of the desired signal in each of plural converted samples using correlation of the desired signal to bodies of known matching samples;
a flagger coupled to the measurer to flag a given converted sample if the confidence for the given converted sample satisfies a target confidence level;
a feedback loop to feed back into the comparator manual inputs that are mislabeled corrections of mis-flagging using inputted text as the flagging for the given converted sample; and
a corrector to re-correct the mislabeled corrections by automatically inputting text as the flagging for the given converted sample, wherein the mislabeled corrections are mislabeled correction of mis-flagging from a human;
wherein each given converted sample comprises a product of a conversion of a respective original sample, the conversion including filtering noise and transforming the original sample from a first form to a second form.
13. The comparator of claim 12 wherein re-correcting includes performing a weighting analysis that determines that the mislabeled corrections are mislabeled or suspicious based on call information of the mislabeled corrections.
14. The comparator of claim 12 further comprising:
a first identifier unit to identify the desired signal to be spotted;
a second identifier unit to identify the target confidence level;
a synchronizer coupled to the first and second identifier units to synchronize a pool of cabined arrays of call samples with the desired signal and the identified confidence level, wherein the cabined arrays are derived from respective distinct environments, the cabined arrays comprising plural converted samples; and
a tester coupled to the synchronizer to pass the pool of cabined arrays of calls samples through the measurer.
15. The comparator of claim 12 further comprising a deriver to derive the bodies of known matching samples from a combination of manually selected samples and samples selected by the comparator.
16. The comparator of claim 12 further comprising an anonymizer to anonymize the converted sample.
17. The comparator of claim 12 wherein the first and second identifying units further comprise a user input device; and
wherein the comparator further comprises a predictive model for performing a weighting analysis comparison of the identified signal to bodies of known matching samples.