echogarden 2.8.7 → 2.9.0
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/dist/api/Recognition.d.ts +3 -3
- package/dist/api/Recognition.js +5 -5
- package/dist/api/Recognition.js.map +1 -1
- package/dist/api/SpeechTranslation.d.ts +4 -4
- package/dist/api/SpeechTranslation.js +4 -4
- package/dist/api/SpeechTranslation.js.map +1 -1
- package/dist/denoising/NSNet2.js +8 -8
- package/dist/denoising/NSNet2.js.map +1 -1
- package/dist/dsp/FFT.d.ts +2 -2
- package/dist/dsp/FFT.js +19 -15
- package/dist/dsp/FFT.js.map +1 -1
- package/dist/dsp/MFCC.d.ts +1 -1
- package/dist/dsp/MFCC.js +8 -8
- package/dist/dsp/MFCC.js.map +1 -1
- package/dist/dsp/{MelSpectogram.d.ts → MelSpectrogram.d.ts} +5 -5
- package/dist/dsp/{MelSpectogram.js → MelSpectrogram.js} +9 -9
- package/dist/dsp/MelSpectrogram.js.map +1 -0
- package/dist/math/VectorMath.d.ts +5 -1
- package/dist/math/VectorMath.js +12 -1
- package/dist/math/VectorMath.js.map +1 -1
- package/dist/nlp/IPA.d.ts +3 -3
- package/dist/recognition/WhisperSTT.d.ts +5 -5
- package/dist/recognition/WhisperSTT.js +38 -32
- package/dist/recognition/WhisperSTT.js.map +1 -1
- package/dist/source-separation/MDXNetSourceSeparation.js +5 -5
- package/dist/source-separation/MDXNetSourceSeparation.js.map +1 -1
- package/dist/speech-embeddings/WavToVec2BertFeatureEmbeddings.js +14 -14
- package/dist/speech-embeddings/WavToVec2BertFeatureEmbeddings.js.map +1 -1
- package/dist/utilities/Utilities.d.ts +1 -1
- package/docs/Tasklist.md +8 -0
- package/package.json +13 -13
- package/src/api/Recognition.ts +10 -7
- package/src/api/SpeechTranslation.ts +16 -6
- package/src/denoising/NSNet2.ts +8 -8
- package/src/dsp/FFT.ts +23 -17
- package/src/dsp/MFCC.ts +8 -8
- package/src/dsp/{MelSpectogram.ts → MelSpectrogram.ts} +9 -8
- package/src/math/VectorMath.ts +17 -1
- package/src/recognition/WhisperSTT.ts +61 -35
- package/src/source-separation/MDXNetSourceSeparation.ts +5 -5
- package/src/speech-embeddings/WavToVec2BertFeatureEmbeddings.ts +14 -14
- package/dist/dsp/MelSpectogram.js.map +0 -1
|
@@ -2,7 +2,7 @@ import type * as Onnx from 'onnxruntime-node'
|
|
|
2
2
|
import { OnnxExecutionProvider, getOnnxSessionOptions } from '../utilities/OnnxUtilities.js';
|
|
3
3
|
|
|
4
4
|
import { RawAudio } from "../audio/AudioUtilities.js";
|
|
5
|
-
import {
|
|
5
|
+
import { computeMelSpectrogram } from "../dsp/MelSpectrogram.js";
|
|
6
6
|
import { Logger } from '../utilities/Logger.js';
|
|
7
7
|
import { concatFloat32Arrays, splitFloat32Array } from '../utilities/Utilities.js';
|
|
8
8
|
import { applyEmphasis } from '../dsp/MFCC.js';
|
|
@@ -31,7 +31,7 @@ export class Wav2Vec2BertFeatureEmbeddings {
|
|
|
31
31
|
|
|
32
32
|
rawAudio.audioChannels[0] = applyEmphasis(rawAudio.audioChannels[0], 0.97)
|
|
33
33
|
|
|
34
|
-
const {
|
|
34
|
+
const { melSpectrogram } = await computeMelSpectrogram(
|
|
35
35
|
rawAudio,
|
|
36
36
|
512,
|
|
37
37
|
400,
|
|
@@ -42,8 +42,8 @@ export class Wav2Vec2BertFeatureEmbeddings {
|
|
|
42
42
|
'povey')
|
|
43
43
|
|
|
44
44
|
// Ensure even length
|
|
45
|
-
if (
|
|
46
|
-
|
|
45
|
+
if (melSpectrogram.length % 2 != 0) {
|
|
46
|
+
melSpectrogram.push(new Float32Array(80))
|
|
47
47
|
}
|
|
48
48
|
|
|
49
49
|
// Normalize filterbanks
|
|
@@ -51,24 +51,24 @@ export class Wav2Vec2BertFeatureEmbeddings {
|
|
|
51
51
|
let sum = 0
|
|
52
52
|
let sumOfSquares = 0
|
|
53
53
|
|
|
54
|
-
for (let i = 0; i <
|
|
55
|
-
const value =
|
|
54
|
+
for (let i = 0; i < melSpectrogram.length; i++) {
|
|
55
|
+
const value = melSpectrogram[i][filterbankIndex]
|
|
56
56
|
|
|
57
57
|
sum += value
|
|
58
58
|
sumOfSquares += value ** 2
|
|
59
59
|
}
|
|
60
60
|
|
|
61
|
-
const mean = sum /
|
|
62
|
-
const normalizationFactor = 1 / (Math.sqrt(sumOfSquares /
|
|
61
|
+
const mean = sum / melSpectrogram.length
|
|
62
|
+
const normalizationFactor = 1 / (Math.sqrt(sumOfSquares / melSpectrogram.length) + 1e-40)
|
|
63
63
|
|
|
64
|
-
for (let i = 0; i <
|
|
65
|
-
|
|
66
|
-
|
|
64
|
+
for (let i = 0; i < melSpectrogram.length; i++) {
|
|
65
|
+
melSpectrogram[i][filterbankIndex] -= mean
|
|
66
|
+
melSpectrogram[i][filterbankIndex] *= normalizationFactor
|
|
67
67
|
}
|
|
68
68
|
}
|
|
69
69
|
|
|
70
70
|
// Flatten
|
|
71
|
-
const
|
|
71
|
+
const flattenedMelSpectrogram = concatFloat32Arrays(melSpectrogram)
|
|
72
72
|
|
|
73
73
|
// Initialize session
|
|
74
74
|
await this.initializeSessionIfNeeded()
|
|
@@ -77,9 +77,9 @@ export class Wav2Vec2BertFeatureEmbeddings {
|
|
|
77
77
|
|
|
78
78
|
const Onnx = await import('onnxruntime-node')
|
|
79
79
|
|
|
80
|
-
const inputTensor = new Onnx.Tensor('float32',
|
|
80
|
+
const inputTensor = new Onnx.Tensor('float32', flattenedMelSpectrogram, [1, melSpectrogram.length / 2, 80 * 2])
|
|
81
81
|
|
|
82
|
-
const attentionMask = new Int32Array(
|
|
82
|
+
const attentionMask = new Int32Array(melSpectrogram.length / 2).fill(1)
|
|
83
83
|
const attentionMaskTensor = new Onnx.Tensor('int32', attentionMask, [1, attentionMask.length])
|
|
84
84
|
|
|
85
85
|
// Run inference
|
|
@@ -1 +0,0 @@
|
|
|
1
|
-
{"version":3,"file":"MelSpectogram.js","sourceRoot":"","sources":["../../src/dsp/MelSpectogram.ts"],"names":[],"mappings":"AACA,OAAO,EAAE,MAAM,EAAE,MAAM,wBAAwB,CAAA;AAC/C,OAAO,KAAK,GAAG,MAAM,UAAU,CAAA;AAE/B,MAAM,CAAC,KAAK,UAAU,oBAAoB,CAAC,QAAkB,EAAE,QAAgB,EAAE,UAAkB,EAAE,SAAiB,EAAE,eAAuB,EAAE,gBAAwB,EAAE,gBAAwB,EAAE,aAA6B,MAAM;IACvO,MAAM,MAAM,GAAG,IAAI,MAAM,EAAE,CAAA;IAE3B,MAAM,CAAC,KAAK,CAAC,wBAAwB,CAAC,CAAA;IACtC,MAAM,QAAQ,GAAG,CAAC,QAAQ,GAAG,CAAC,CAAC,GAAG,CAAC,CAAA;IACnC,MAAM,gBAAgB,GAAG,QAAQ,CAAC,UAAU,GAAG,CAAC,CAAA;IAChD,MAAM,cAAc,GAAG,GAAG,CAAC,iBAAiB,CAAC,QAAQ,EAAE,gBAAgB,CAAC,CAAA;IAExE,MAAM,iBAAiB,GAAG,UAAU,CAAC,gBAAgB,CAAC,CAAA;IACtD,MAAM,iBAAiB,GAAG,UAAU,CAAC,gBAAgB,CAAC,CAAA;IAEtD,MAAM,4BAA4B,GAAG,kCAAkC,CAAC,eAAe,EAAE,iBAAiB,EAAE,iBAAiB,CAAC,CAAA;IAC9H,MAAM,cAAc,GAAG,iBAAiB,CAAC,cAAc,EAAE,4BAA4B,EAAE,iBAAiB,EAAE,iBAAiB,CAAC,CAAA;IAE5H,MAAM,CAAC,GAAG,EAAE,CAAA;IAEZ,OAAO,oCAAoC,CAAC,QAAQ,EAAE,QAAQ,EAAE,UAAU,EAAE,SAAS,EAAE,cAAc,EAAE,UAAU,CAAC,CAAA;AACnH,CAAC;AAED,MAAM,CAAC,KAAK,UAAU,oCAAoC,CAAC,QAAkB,EAAE,QAAgB,EAAE,UAAkB,EAAE,SAAiB,EAAE,WAAyB,EAAE,aAA6B,MAAM;IACrM,MAAM,MAAM,GAAG,IAAI,MAAM,EAAE,CAAA;IAE3B,MAAM,CAAC,KAAK,CAAC,yBAAyB,CAAC,CAAA;IACvC,MAAM,YAAY,GAAG,QAAQ,CAAC,aAAa,CAAC,CAAC,CAAC,CAAA;IAC9C,MAAM,SAAS,GAAG,MAAM,GAAG,CAAC,KAAK,CAAC,YAAY,EAAE,QAAQ,EAAE,UAAU,EAAE,SAAS,EAAE,UAAU,CAAC,CAAA;IAE5F,MAAM,CAAC,KAAK,CAAC,wCAAwC,CAAC,CAAA;IACtD,MAAM,aAAa,GAAG,wBAAwB,CAAC,SAAS,EAAE,WAAW,CAAC,CAAA;IAEtE,MAAM,CAAC,GAAG,EAAE,CAAA;IAEZ,OAAO,EAAE,aAAa,EAAE,SAAS,EAAE,CAAA;AACpC,CAAC;AAED,MAAM,UAAU,wBAAwB,CAAC,SAAyB,EAAE,cAA4B;IAC/F,OAAO,SAAS,CAAC,GAAG,CAAC,QAAQ,CAAC,EAAE;QAC/B,MAAM,aAAa,GAAG,GAAG,CAAC,uBAAuB,CAAC,QAAQ,CAAC,CAAA;QAC3D,OAAO,0BAA0B,CAAC,aAAa,EAAE,cAAc,CAAC,CAAA;IACjE,CAAC,CAAC,CAAA;AACH,CAAC;AAED,MAAM,UAAU,0BAA0B,CAAC,aAA2B,EAAE,WAAyB;IAChG,MAAM,eAAe,GAAG,WAAW,CAAC,MAAM,CAAA;IAC1C,MAAM,WAAW,GAAG,IAAI,YAAY,CAAC,eAAe,CAAC,CAAA;IAErD,KAAK,IAAI,YAAY,GAAG,CAAC,EAAE,YAAY,GAAG,eAAe,EAAE,YAAY,EAAE,EAAE,CAAC;QAC3E,MAAM,UAAU,GAAG,WAAW,CAAC,YAAY,CAAC,CAAA;QAC5C,MAAM,oBAAoB,GAAG,UAAU,CAAC,UAAU,CAAA;QAClD,MAAM,iBAAiB,GAAG,UAAU,CAAC,OAAO,CAAA;QAE5C,IAAI,oBAAoB,KAAK,CAAC,CAAC,EAAE,CAAC;YACjC,SAAQ;QACT,CAAC;QAED,IAAI,YAAY,GAAG,CAAC,CAAA;QAEpB,KAAK,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,GAAG,iBAAiB,CAAC,MAAM,EAAE,CAAC,EAAE,EAAE,CAAC;YACnD,MAAM,kBAAkB,GAAG,oBAAoB,GAAG,CAAC,CAAA;YAEnD,IAAI,kBAAkB,IAAI,aAAa,CAAC,MAAM,EAAE,CAAC;gBAChD,MAAK;YACN,CAAC;YAED,MAAM,MAAM,GAAG,iBAAiB,CAAC,CAAC,CAAC,CAAA;YACnC,MAAM,kBAAkB,GAAG,aAAa,CAAC,kBAAkB,CAAC,CAAA;YAE5D,YAAY,IAAI,MAAM,GAAG,kBAAkB,CAAA;QAC5C,CAAC;QAED,WAAW,CAAC,YAAY,CAAC,GAAG,YAAY,CAAA;IACzC,CAAC;IAED,OAAO,WAAW,CAAA;AACnB,CAAC;AAED,MAAM,UAAU,iBAAiB,CAAC,0BAAwC,EAAE,oBAAkC,EAAE,iBAAyB,EAAE,iBAAyB;IACnK,MAAM,eAAe,GAAG,oBAAoB,CAAC,MAAM,CAAA;IACnD,MAAM,2BAA2B,GAAG,0BAA0B,CAAC,GAAG,CAAC,WAAW,CAAC,EAAE,CAAC,UAAU,CAAC,WAAW,CAAC,CAAC,CAAA;IAE1G,MAAM,WAAW,GAAiB,EAAE,CAAA;IAEpC,KAAK,IAAI,eAAe,GAAG,CAAC,EAAE,eAAe,GAAG,eAAe,EAAE,eAAe,EAAE,EAAE,CAAC;QACpF,MAAM,eAAe,GAAG,oBAAoB,CAAC,eAAe,CAAC,CAAA;QAE7D,MAAM,aAAa,GAAG,eAAe,GAAG,CAAC,CAAC,CAAC,CAAC,oBAAoB,CAAC,eAAe,GAAG,CAAC,CAAC,CAAC,CAAC,CAAC,iBAAiB,CAAA;QACzG,MAAM,cAAc,GAAG,eAAe,GAAG,eAAe,GAAG,CAAC,CAAC,CAAC,CAAC,oBAAoB,CAAC,eAAe,GAAG,CAAC,CAAC,CAAC,CAAC,CAAC,iBAAiB,CAAA;QAE5H,MAAM,KAAK,GAAG,cAAc,GAAG,aAAa,CAAA;QAC5C,MAAM,SAAS,GAAG,KAAK,GAAG,CAAC,CAAA;QAE3B,IAAI,UAAU,GAAG,CAAC,CAAC,CAAA;QACnB,IAAI,OAAO,GAAa,EAAE,CAAA;QAE1B,IAAI,SAAS,GAAG,CAAC,CAAA;QAEjB,KAAK,IAAI,sBAAsB,GAAG,CAAC,EAAE,sBAAsB,GAAG,2BAA2B,CAAC,MAAM,EAAE,sBAAsB,EAAE,EAAE,CAAC;YAC5H,MAAM,6BAA6B,GAAG,2BAA2B,CAAC,sBAAsB,CAAC,CAAA;YAEzF,IAAI,MAAM,GAAG,CAAC,CAAA;YAEd,IAAI,6BAA6B,IAAI,aAAa,IAAI,6BAA6B,IAAI,eAAe,EAAE,CAAC;gBACxG,MAAM,GAAG,CAAC,6BAA6B,GAAG,aAAa,CAAC,GAAG,SAAS,CAAA;YACrE,CAAC;iBAAM,IAAI,6BAA6B,GAAG,eAAe,IAAI,6BAA6B,IAAI,cAAc,EAAE,CAAC;gBAC/G,MAAM,GAAG,CAAC,cAAc,GAAG,6BAA6B,CAAC,GAAG,SAAS,CAAA;YACtE,CAAC;YAED,IAAI,MAAM,GAAG,CAAC,EAAE,CAAC;gBAChB,IAAI,UAAU,IAAI,CAAC,CAAC,EAAE,CAAC;oBACtB,UAAU,GAAG,sBAAsB,CAAA;gBACpC,CAAC;gBAED,OAAO,CAAC,IAAI,CAAC,MAAM,CAAC,CAAA;gBACpB,SAAS,IAAI,MAAM,CAAA;YACpB,CAAC;iBAAM,IAAI,UAAU,IAAI,CAAC,CAAC,EAAE,CAAC;gBAC7B,MAAK;YACN,CAAC;QACF,CAAC;QAED,OAAO,GAAG,OAAO,CAAC,GAAG,CAAC,MAAM,CAAC,EAAE,CAAC,MAAM,GAAG,SAAS,CAAC,CAAA;QAEnD,WAAW,CAAC,IAAI,CAAC,EAAE,UAAU,EAAE,OAAO,EAAE,CAAC,CAAA;IAC1C,CAAC;IAED,OAAO,WAAW,CAAA;AACnB,CAAC;AAED,MAAM,UAAU,kCAAkC,CAAC,YAAoB,EAAE,iBAAyB,EAAE,iBAAyB;IAC5H,MAAM,WAAW,GAAG,CAAC,iBAAiB,GAAG,iBAAiB,CAAC,GAAG,CAAC,YAAY,GAAG,CAAC,CAAC,CAAA;IAEhF,MAAM,iBAAiB,GAAG,IAAI,YAAY,CAAC,YAAY,CAAC,CAAA;IAExD,KAAK,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,GAAG,YAAY,EAAE,CAAC,EAAE,EAAE,CAAC;QACvC,iBAAiB,CAAC,CAAC,CAAC,GAAG,iBAAiB,GAAG,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,GAAG,WAAW,CAAC,CAAA;IACnE,CAAC;IAED,OAAO,iBAAiB,CAAA;AACzB,CAAC;AAED,MAAM,UAAU,UAAU,CAAC,SAAiB;IAC3C,OAAO,MAAM,GAAG,IAAI,CAAC,KAAK,CAAC,GAAG,GAAG,CAAC,SAAS,GAAG,KAAK,CAAC,CAAC,CAAA;AACtD,CAAC;AAED,MAAM,UAAU,UAAU,CAAC,GAAW;IACrC,OAAO,KAAK,GAAG,CAAC,IAAI,CAAC,GAAG,CAAC,IAAI,EAAE,GAAG,GAAG,MAAM,CAAC,GAAG,GAAG,CAAC,CAAA;AACpD,CAAC"}
|