| { |
| "field_id": "F019", |
| "field": "audio machine learning", |
| "directions": [ |
| { |
| "direction_id": "G0", |
| "name": "Audio authenticity and adversarial security", |
| "scope": "Study how forged or intentionally crafted audio affects audio-based judgments, including detecting, locating, and attributing synthesized or manipulated content, as well as attacks on audio recognition systems and defenses against them." |
| }, |
| { |
| "direction_id": "G1", |
| "name": "Affective inference from audio", |
| "scope": "Infer expressed or perceived emotions and other affective attributes from speech, vocalizations, or music." |
| }, |
| { |
| "direction_id": "G2", |
| "name": "Acoustic content recognition across categories and evidence conditions", |
| "scope": "Identify environmental or meaningful sound categories in recordings and, when the task requires it, determine event times; also recognize insufficiently seen categories using semantic descriptions or a small number of sound examples." |
| }, |
| { |
| "direction_id": "G3", |
| "name": "Reusable and resource-conscious audio recognition models", |
| "scope": "Study general-purpose models and frontends that support audio recognition by providing useful representations across downstream tasks or preserving recognition performance under resource constraints." |
| }, |
| { |
| "direction_id": "G4", |
| "name": "Musical content and performance analysis", |
| "scope": "Infer musical attributes or performance structure from music and performed sounds, including categorical descriptions and time-varying content such as pitch, notes, and percussion events." |
| }, |
| { |
| "direction_id": "G5", |
| "name": "Recovery of constituent sounds from mixtures", |
| "scope": "Recover constituent sounds from mixed recordings, whether an external target condition specifies the source to return or multiple sources are decomposed without a specified target." |
| }, |
| { |
| "direction_id": "G6", |
| "name": "Spatial inference and transfer of sound", |
| "scope": "Study the relationship between sound and space by inferring the positions and movements of sounding sources or estimating and rendering how rooms and source–receiver geometry transform sound propagation." |
| }, |
| { |
| "direction_id": "G7", |
| "name": "Recovery of degraded existing recordings", |
| "scope": "Restore desired sound using an existing recording as evidence, both by suppressing interference such as noise and reverberation and by supplying missing content when segments are absent, bandwidth is insufficient, or damage is severe." |
| }, |
| { |
| "direction_id": "G8", |
| "name": "Audio synthesis, rendering and voice transformation", |
| "scope": "Produce or transform audible sound by creating sounds or music from high-level conditions, rendering waveforms from acoustic representations, designing or resynthesizing timbre through controllable parameters, or converting existing spoken or sung content into a target voice or style." |
| }, |
| { |
| "direction_id": "G9", |
| "name": "Linguistic inference from speech", |
| "scope": "Infer linguistic information from spoken audio, including transcribing the words spoken and identifying or comparing the language, dialect, or accent variety used." |
| }, |
| { |
| "direction_id": "G10", |
| "name": "Speaker attribution in recordings", |
| "scope": "Attribute speech to speakers using audio or audiovisual evidence, both by identifying or verifying an individual’s identity and by determining when each person speaks in a recording." |
| }, |
| { |
| "direction_id": "G11", |
| "name": "Audio-based human health and functional assessment", |
| "scope": "Use recordings of sounds produced by people as evidence of health or functional status to assess respiratory diseases and abnormal patterns, or speech disorders, fluency, and psychological distress." |
| }, |
| { |
| "direction_id": "G12", |
| "name": "Machine condition monitoring from sound", |
| "scope": "Detect abnormal operation or diagnose mechanical faults from the sounds of operating machines." |
| }, |
| { |
| "direction_id": "G13", |
| "name": "Audio–language description and retrieval", |
| "scope": "Establish correspondences between recordings and descriptive language so that users can find recordings with queries or generate natural-language descriptions from recordings." |
| }, |
| { |
| "direction_id": "G14", |
| "name": "Perceptual assessment of audio", |
| "scope": "Predict or systematically evaluate listeners’ perceived audio or speech quality, naturalness, intelligibility, or preference, rather than primarily producing improved sound." |
| } |
| ] |
| } |
|
|
