Uses 911K of libraries. Click here for Pure Java version (4910L/25K).
!7 // Note: This is dependent on input buffer size which is not optimal. // Tested with buffer size 232 ms (using CD quality) cmodule VAD > DynPrintLogAndEnabled { switchable double windowSize = 1; // milliseconds switchable double threshold = // 10.0; // the magic value for 44 KHz stereo 1; // for 8 KHz mono transient double confidence; transient short[] fromData; // only useful when sendOutClonedData = true start { printLogUpdateInterval = 100; // technically we are treating stereo input as mono samples // but it might not matter dm_addAudioListener(voidfunc(short[] data) enter { if (!enabled) ret; short[] buffer = cloneArray(data); dm_q(r { if (q.size() > 2) ret with print("overload"); short[] buf = buffer; //printStruct(takeFirstOfShortArray(8, buf)); float sampleRate = dm_audioInputSampleRate(); if (sampleRate == 0) ret; float originalSampleRate = sampleRate; int channels = dm_audioInputChannels(); if (channels > 1) buf = mixStereoShortArrayToMono(buf); buf = convertSampleRate_shortArray_simple(buf, sampleRate, 8000); sampleRate = 8000f; long time = sysNow(); int windowLength = msToSamples_int(windowSize, sampleRate); //printWithMilliseconds("Window length (" + sampleRate + "): " + windowLength); new L<Double> l; L<IntRange> chunks = intRangeChunks(buf, windowLength); for (IntRange r : chunks) l.add(normalizedAutocorrelationOfAudioSamples(buf, r)); // print one table for reference print(asList(normalizedAutocorrelationTableForAudioSamples(buf, first(chunks)))); double result = doubleAverage(l); confidence = result/threshold*50; fromData = data; int vol = shortAbsMax(buf); print("VAD " + (confidence >= 50 ? "XXX" : " ") + " value: " + iround(confidence) + " % [" + elapsedMS(time) + " ms, " + nWindows(l(buf)/windowLength) + " of size " + windowLength + ", converted from \*originalSampleRate*/ Hz, " + nChannels(channels) + ", volume: " + iround(vol/32767.0*100) + "]"); }); }); } // API Bool hasVoiceActivity(short[] fromData) { ret fromData != this.fromData ? null : confidence >= 50; } }
Began life as a copy of #1026891
download show line numbers debug dex old transpilations
Travelled to 7 computer(s): bhatertpkbcr, mqqgnosmbjvj, pyentgdyhuwx, pzhvpgtvlbxg, tvejysmllsmz, vouqrxazstgt, xrpafgyirdlv
No comments. add comment
Snippet ID: | #1026900 |
Snippet name: | Auto-Correlation VAD Test 1 [works for main mic & normal background noise] |
Eternal ID of this version: | #1026900/28 |
Text MD5: | 2a4baf824dc5ed452578fb6ec752b6a6 |
Transpilation MD5: | 0e135ae5578179ecee8fd2331cdd54ee |
Author: | stefan |
Category: | javax / audio |
Type: | JavaX source code (Dynamic Module) |
Public (visible to everyone): | Yes |
Archived (hidden from active list): | No |
Created/modified: | 2020-02-02 20:44:07 |
Source code size: | 2403 bytes / 60 lines |
Pitched / IR pitched: | No / No |
Views / Downloads: | 301 / 1812 |
Version history: | 27 change(s) |
Referenced in: | -