/**
 * Aggregate (numChunks, numFrames, K) sliding-window predictions onto a global
 * frame timeline. Mimics pyannote.audio.Inference.aggregate(skip_average=False),
 * with a configurable `weighting` callback (default: uniform).
 *
 * Each chunk c starts at chunkStarts[c] seconds, has `numFrames` frames spanning
 * `chunkDurationSec` seconds. The output is a (totalFrames, K) array on a
 * timeline shared with all chunks (frame step inferred from chunkStep).
 */
export interface AggregationResult {
    data: Float64Array;
    totalFrames: number;
    numChannels: number;
    /** Frame duration in seconds. */
    frameDurationSec: number;
    /** Number of overlapping chunks contributing to each frame (Float64Array). */
    weights: Float64Array;
}
export declare function aggregateSlidingWindows(data: Uint8Array | Float32Array | Float64Array, numChunks: number, numFramesPerChunk: number, numChannels: number, chunkStartsSec: Float64Array, chunkDurationSec: number): AggregationResult;
/** Frame-level instantaneous speaker count from binarized segmentations. */
export declare function speakerCount(segData: Uint8Array, numChunks: number, numFrames: number, numLocalSpeakers: number, chunkStartsSec: Float64Array, chunkDurationSec: number): {
    count: Uint8Array;
    totalFrames: number;
    frameDurationSec: number;
};
//# sourceMappingURL=aggregate.d.ts.map