Neutralizing distortion in audio data

A system for neutralizing distortion in audio data comprising:

• one or more memories;
• and
• at least one processor of a transmitting device coupled to the one or more memories, and configured to:
• produce candidate audio signals by applying distortions to initial audio signals;
• process the candidate audio signals to produce resulting audio signals representing the candidate audio signals after transmission to a receiving device;
• generate transcriptions of the initial audio signals and the resulting audio signals;
• determine intelligibility scores for the transcriptions of the initial audio signals and the resulting audio signals;
• generate a training set by selecting the initial audio signals and corresponding resulting audio signals based on the intelligibility scores for the transcriptions;
• train a machine learning model to produce a distorted audio signal based on the training set;
• process, via the machine learning model, an audio signal from an audio portion of a media signal containing media to produce the distorted audio signal, wherein the distorted audio signal compensates for distortion from transmission of the audio signal to the receiving device and provides greater intelligibility relative to the audio signal;
• and
• enable transmission of the distorted audio signal produced from the media signal to the receiving device for presentation of the media.

A computer program product for neutralizing distortion in audio data, the computer program product comprising one or more non-transitory computer readable storage media having program instructions stored thereon, the program instructions executable by at least one processor of a transmitting device to cause the at least one processor to:

• produce candidate audio signals by applying distortions to initial audio signals;
• process the candidate audio signals to produce resulting audio signals representing the candidate audio signals after transmission to a receiving device;
• generate transcriptions of the initial audio signals and the resulting audio signals;
• determine intelligibility scores for the transcriptions of the initial audio signals and the resulting audio signals;
• generate a training set by selecting the initial audio signals and corresponding resulting audio signals based on the intelligibility scores for the transcriptions;
• train a machine learning model to produce a distorted audio signal based on the training set;
• process, via the machine learning model, an audio signal from an audio portion of a media signal containing media to produce the distorted audio signal, wherein the distorted audio signal compensates for distortion from transmission of the audio signal to the receiving device and provides greater intelligibility relative to the audio signal;
• and
• enable transmission of the distorted audio signal produced from the media signal to the receiving device for presentation of the media.