Initial Commit

packages/media/audio/audio-analysis-spectrum/src/analyzer/spectrogram.ts

@@ -0,0 +1,103 @@
+import type { AudioDocument } from "@suaveplan/audio-core/document/document";
+import { frameCount, frameMagnitude, readMono } from "./dsp.js";
+import type { SpectrumAnalyzerOptionsInput } from "./options.schemas.js";
+import type { SpectrogramResult } from "./result.types.js";
+import {
+    buildBase,
+    ensureNotAborted,
+    parseOptions,
+    startRun,
+} from "./runtime.js";
+
+export interface SpectrogramRunOptions extends SpectrumAnalyzerOptionsInput {
+    readonly signal?: AbortSignal;
+}
+
+function emptyResult(doc: AudioDocument): SpectrogramResult {
+    return {
+        ...buildBase("spectrogram", doc),
+        op: "spectrogram",
+        spectrogram: [],
+    };
+}
+
+export class Spectrogram {
+    public readonly name = "spectrogram";
+    public readonly description =
+        "Time-frequency magnitude matrix of an audio document.";
+
+    public async analyze(
+        doc: AudioDocument,
+        raw?: SpectrogramRunOptions
+    ): Promise<SpectrogramResult> {
+        const opts = parseOptions(raw);
+        ensureNotAborted(raw?.signal);
+        const handle = startRun("spectrogram", doc, "analyze");
+        try {
+            const samples = await readMono(doc);
+            const count = frameCount(
+                samples.length,
+                opts.frameSize,
+                opts.hopSize
+            );
+            const spectrogram: Float32Array[] = [];
+            for (let i = 0; i < count; i++) {
+                ensureNotAborted(raw?.signal);
+                const offset = i * opts.hopSize;
+                spectrogram.push(
+                    frameMagnitude(samples, offset, opts.frameSize, opts.window)
+                );
+            }
+            handle.end();
+            return {
+                ...buildBase("spectrogram", doc),
+                op: "spectrogram",
+                spectrogram,
+            };
+        } catch (err) {
+            handle.fail(err);
+            throw err;
+        }
+    }
+
+    public async *analyzeStream(
+        doc: AudioDocument,
+        raw?: SpectrogramRunOptions
+    ): AsyncGenerator<SpectrogramResult> {
+        const opts = parseOptions(raw);
+        ensureNotAborted(raw?.signal);
+        const handle = startRun("spectrogram", doc, "stream");
+        try {
+            const samples = await readMono(doc);
+            const count = frameCount(
+                samples.length,
+                opts.frameSize,
+                opts.hopSize
+            );
+            if (count === 0) {
+                yield emptyResult(doc);
+                handle.end();
+                return;
+            }
+            for (let i = 0; i < count; i++) {
+                ensureNotAborted(raw?.signal);
+                const offset = i * opts.hopSize;
+                const mag = frameMagnitude(
+                    samples,
+                    offset,
+                    opts.frameSize,
+                    opts.window
+                );
+                yield {
+                    ...buildBase("spectrogram", doc),
+                    op: "spectrogram",
+                    spectrogram: [mag],
+                };
+            }
+            handle.end();
+        } catch (err) {
+            handle.fail(err);
+            throw err;
+        }
+    }
+}