mirror of
https://github.com/vale981/VoiDPlugins
synced 2025-03-04 16:51:38 -05:00
Add ExpASFilter
It uses ML techniques to process tablet input and achieve three objectives depending on configuration: smoothing, latency compensation, and noise reduction.
This commit is contained in:
X9VoiD
parent
0543da0a9f
commit
ec881bd0c5
3 changed files with 346 additions and 0 deletions
322
ExperimentalASFilter/ExperimentalASFilter.cs
Normal file
322
ExperimentalASFilter/ExperimentalASFilter.cs
Normal file
|
|
@ -0,0 +1,322 @@
|
|||
using ExpASFilter;
|
||||
using MathNet.Numerics;
|
||||
using MathNet.Numerics.LinearRegression;
|
||||
using System;
|
||||
using System.Collections.Generic;
|
||||
using TabletDriverPlugin;
|
||||
using TabletDriverPlugin.Attributes;
|
||||
using TabletDriverPlugin.Tablet;
|
||||
|
||||
namespace OpenTabletDriverPlugins
|
||||
{
|
||||
|
||||
[PluginName("Experimental AS Filter")]
|
||||
public class ExperimentalASFilter : Notifier, IFilter
|
||||
{
|
||||
public virtual Point Filter(Point point)
|
||||
{
|
||||
DateTime date = DateTime.Now;
|
||||
CalcReportRate(date);
|
||||
var reportRateAvg = CalcReportRateAvg();
|
||||
|
||||
if (AddTimeSeriesPoint(point, date))
|
||||
{
|
||||
var predicted = new Point();
|
||||
var timeMatrix = ConstructTimeDesignMatrix();
|
||||
double[] x, y;
|
||||
if (Normalize)
|
||||
{
|
||||
x = ConstructNormalizedTargetMatrix(Axis.X);
|
||||
y = ConstructNormalizedTargetMatrix(Axis.Y);
|
||||
}
|
||||
else
|
||||
{
|
||||
x = ConstructTargetMatrix(Axis.X);
|
||||
y = ConstructTargetMatrix(Axis.Y);
|
||||
}
|
||||
|
||||
Polynomial xCoeff, yCoeff;
|
||||
if (Weighted)
|
||||
{
|
||||
var weights = CalcWeight();
|
||||
xCoeff = new Polynomial(Fit.PolynomialWeighted(timeMatrix, x, weights, Degree));
|
||||
yCoeff = new Polynomial(Fit.PolynomialWeighted(timeMatrix, y, weights, Degree));
|
||||
}
|
||||
else
|
||||
{
|
||||
xCoeff = new Polynomial(Fit.Polynomial(timeMatrix, x, Degree, DirectRegressionMethod.Svd));
|
||||
yCoeff = new Polynomial(Fit.Polynomial(timeMatrix, y, Degree, DirectRegressionMethod.Svd));
|
||||
}
|
||||
|
||||
double predictAhead;
|
||||
if (Sync)
|
||||
predictAhead = 1000.0 / reportRateAvg * Ahead;
|
||||
else
|
||||
predictAhead = (date - _timeSeriesPoints.First.Value.Date).TotalMilliseconds + Compensation;
|
||||
|
||||
predicted.X = (float)xCoeff.Evaluate(predictAhead);
|
||||
predicted.Y = (float)yCoeff.Evaluate(predictAhead);
|
||||
if (Normalize)
|
||||
{
|
||||
predicted.X *= ScreenWidth;
|
||||
predicted.Y *= ScreenHeight;
|
||||
}
|
||||
|
||||
var now = DateTime.Now;
|
||||
if ((now - date).TotalMilliseconds > CalcReportRateAvg())
|
||||
Log.Write("ExpASFilter", now + ": CPU choking hard. Latency higher than normal. We missed a hz.");
|
||||
|
||||
_lastTime = date;
|
||||
return predicted;
|
||||
}
|
||||
_lastTime = date;
|
||||
return point;
|
||||
}
|
||||
|
||||
#region Private Functions
|
||||
|
||||
private bool AddTimeSeriesPoint(Point point, DateTime time)
|
||||
{
|
||||
_timeSeriesPoints.AddLast(new TimeSeriesPoint(point, time));
|
||||
if (_timeSeriesPoints.Count > Samples)
|
||||
_timeSeriesPoints.RemoveFirst();
|
||||
if (_timeSeriesPoints.Count == Samples)
|
||||
return true;
|
||||
return false;
|
||||
}
|
||||
|
||||
private double[] ConstructTimeDesignMatrix()
|
||||
{
|
||||
DateTime baseTime = _timeSeriesPoints.First.Value.Date;
|
||||
|
||||
var data = new double[Samples];
|
||||
var index = -1;
|
||||
foreach (var timePoint in _timeSeriesPoints)
|
||||
{
|
||||
++index;
|
||||
data[index] = (timePoint.Date - baseTime).TotalMilliseconds;
|
||||
}
|
||||
|
||||
return data;
|
||||
}
|
||||
|
||||
private double[] ConstructTargetMatrix(Axis axis)
|
||||
{
|
||||
var points = new double[Samples];
|
||||
var index = -1;
|
||||
|
||||
if (axis == Axis.X)
|
||||
foreach (var timePoint in _timeSeriesPoints)
|
||||
{
|
||||
points[++index] = timePoint.Point.X;
|
||||
}
|
||||
|
||||
else if (axis == Axis.Y)
|
||||
foreach (var timePoint in _timeSeriesPoints)
|
||||
{
|
||||
points[++index] = timePoint.Point.Y;
|
||||
}
|
||||
|
||||
return points;
|
||||
}
|
||||
|
||||
private double[] ConstructNormalizedTargetMatrix(Axis axis)
|
||||
{
|
||||
var points = new double[Samples];
|
||||
var index = -1;
|
||||
|
||||
if (axis == Axis.X)
|
||||
foreach (var timePoint in _timeSeriesPoints)
|
||||
{
|
||||
points[++index] = timePoint.Point.X / ScreenWidth;
|
||||
}
|
||||
|
||||
else if (axis == Axis.Y)
|
||||
foreach (var timePoint in _timeSeriesPoints)
|
||||
{
|
||||
points[++index] = timePoint.Point.Y / ScreenHeight;
|
||||
}
|
||||
|
||||
return points;
|
||||
}
|
||||
|
||||
private void CalcReportRate(DateTime now)
|
||||
{
|
||||
_reportRate = 1000.0 / (now - _lastTime).TotalMilliseconds;
|
||||
_reportRateAvg.AddLast(_reportRate);
|
||||
if (_reportRateAvg.Count > 10)
|
||||
_reportRateAvg.RemoveFirst();
|
||||
}
|
||||
|
||||
private double CalcReportRateAvg()
|
||||
{
|
||||
double avg = 0;
|
||||
foreach (var sample in _reportRateAvg)
|
||||
avg += sample;
|
||||
return avg / _reportRateAvg.Count;
|
||||
}
|
||||
|
||||
private double[] CalcWeight()
|
||||
{
|
||||
var weights = new List<double>();
|
||||
var weightsNormalized = new List<double>();
|
||||
double weight = 1;
|
||||
foreach (var point in _timeSeriesPoints)
|
||||
weights.Add(weight *= 2);
|
||||
foreach (var _weight in weights)
|
||||
weightsNormalized.Add(_weight / weights[^1]);
|
||||
return weightsNormalized.ToArray();
|
||||
}
|
||||
|
||||
#endregion Private Functions
|
||||
|
||||
private int _samples = 4, _degree = 1, _ahead = 6;
|
||||
private LinkedList<TimeSeriesPoint> _timeSeriesPoints = new LinkedList<TimeSeriesPoint>();
|
||||
private LinkedList<double> _reportRateAvg = new LinkedList<double>();
|
||||
private double _reportRate;
|
||||
private DateTime _lastTime = DateTime.Now;
|
||||
private double _compensation;
|
||||
private bool _normalize, _sync, _weighted;
|
||||
private int _screenWidth, _screenHeight;
|
||||
|
||||
#region Controls
|
||||
|
||||
[UnitProperty("Compensation", "ms")]
|
||||
public double Compensation
|
||||
{
|
||||
set => CompensationFunc(ref _compensation, value);
|
||||
get => _compensation;
|
||||
}
|
||||
|
||||
[Property("Samples")]
|
||||
public int Samples
|
||||
{
|
||||
set => SamplesFunc(ref _samples, value);
|
||||
get => _samples;
|
||||
}
|
||||
|
||||
[Property("Degree")]
|
||||
public int Degree
|
||||
{
|
||||
set => DegreeFunc(ref _degree, value);
|
||||
get => _degree;
|
||||
}
|
||||
|
||||
[BooleanProperty("Normalize", "Preprocess the input before performing regression. (may improve accuracy) Set Screen Dimensions below when enabling Normalization.")]
|
||||
public bool Normalize
|
||||
{
|
||||
set => RaiseAndSetIfChanged(ref _normalize, value);
|
||||
get => _normalize;
|
||||
}
|
||||
|
||||
[UnitProperty("Screen Width", "px")]
|
||||
public int ScreenWidth
|
||||
{
|
||||
set => RaiseAndSetIfChanged(ref _screenWidth, value);
|
||||
get => _screenWidth;
|
||||
}
|
||||
|
||||
[UnitProperty("Screen Height", "px")]
|
||||
public int ScreenHeight
|
||||
{
|
||||
set => RaiseAndSetIfChanged(ref _screenHeight, value);
|
||||
get => _screenHeight;
|
||||
}
|
||||
|
||||
[BooleanProperty("Synchronize to Report Rate", "Compensation will be ignored in respect for tablet's report rate. Set below how many tablet reports to predict ahead.")]
|
||||
public bool Sync
|
||||
{
|
||||
set => RaiseAndSetIfChanged(ref _sync, value);
|
||||
get => _sync;
|
||||
}
|
||||
|
||||
[Property("Reports Ahead")]
|
||||
public int Ahead
|
||||
{
|
||||
set => RaiseAndSetIfChanged(ref _ahead, value);
|
||||
get => _ahead;
|
||||
}
|
||||
|
||||
[BooleanProperty("Exponential Weighted Polynomial Regression", "Well...")]
|
||||
public bool Weighted
|
||||
{
|
||||
set => RaiseAndSetIfChanged(ref _weighted, value);
|
||||
get => _weighted;
|
||||
}
|
||||
|
||||
#endregion Controls
|
||||
|
||||
#region Control Utility Functions
|
||||
|
||||
private void CompensationFunc(ref double a, double value)
|
||||
{
|
||||
if (value > 30)
|
||||
Log.Write("ExpASFilter", "Unrealistic latency compensation. [Compensation: " + value + "ms]", true);
|
||||
RaiseAndSetIfChanged(ref a, value);
|
||||
}
|
||||
|
||||
private void SamplesFunc(ref int a, int value)
|
||||
{
|
||||
int minimum = Degree + 1;
|
||||
var suggested = Math.Pow(Degree, 2);
|
||||
|
||||
if (value > 12)
|
||||
Log.Write("ExpASFilter",
|
||||
"Samples too high. Expect higher latency." +
|
||||
"[Samples: " + value + "]", true);
|
||||
|
||||
else if (value <= minimum)
|
||||
{
|
||||
Log.Write("ExpASFilter",
|
||||
"Samples too low for selected degree." +
|
||||
"[Samples: " + value + ", Required Minimum: " + minimum + "]", true);
|
||||
RaiseAndSetIfChanged(ref a, minimum);
|
||||
return;
|
||||
}
|
||||
|
||||
if (value <= suggested)
|
||||
Log.Write("ExpASFilter",
|
||||
"Samples might not be enough for an accurate prediction." +
|
||||
"[Samples: " + value + ", Suggestion: " + suggested + "]");
|
||||
|
||||
RaiseAndSetIfChanged(ref a, value);
|
||||
}
|
||||
|
||||
private void DegreeFunc(ref int a, int value)
|
||||
{
|
||||
if (value == 0)
|
||||
{
|
||||
Log.Write("ExpASFilter", "Degree cannot be zero", true);
|
||||
}
|
||||
else if (value > 5)
|
||||
Log.Write("ExpASFilter", "Degree too high, might cause instability and inaccuracy issues" +
|
||||
"[Suggestion: Degree <= 5]");
|
||||
RaiseAndSetIfChanged(ref a, value);
|
||||
}
|
||||
|
||||
#endregion Control Utility Functions
|
||||
|
||||
public FilterStage FilterStage => FilterStage.PostTranspose;
|
||||
|
||||
}
|
||||
}
|
||||
|
||||
namespace ExpASFilter
|
||||
{
|
||||
enum Axis {
|
||||
X,
|
||||
Y
|
||||
}
|
||||
public class TimeSeriesPoint
|
||||
{
|
||||
public TimeSeriesPoint(Point point, DateTime date)
|
||||
{
|
||||
Point = point;
|
||||
Date = date;
|
||||
}
|
||||
|
||||
public Point Point { get; }
|
||||
public DateTime Date { get; }
|
||||
|
||||
}
|
||||
}
|
||||
18
ExperimentalASFilter/ExperimentalASFilter.csproj
Normal file
18
ExperimentalASFilter/ExperimentalASFilter.csproj
Normal file
|
|
@ -0,0 +1,18 @@
|
|||
<Project Sdk="Microsoft.NET.Sdk">
|
||||
|
||||
<PropertyGroup>
|
||||
<TargetFrameworks>netcoreapp3.1; net5.0</TargetFrameworks>
|
||||
</PropertyGroup>
|
||||
|
||||
<PropertyGroup Condition="'$(Configuration)' == 'Release'">
|
||||
<DebugType>none</DebugType>
|
||||
<DebugSymbols>false</DebugSymbols>
|
||||
<CopyOutputSymbolsToOutputDirectory>false</CopyOutputSymbolsToOutputDirectory>
|
||||
</PropertyGroup>
|
||||
|
||||
<ItemGroup>
|
||||
<PackageReference Include="MathNet.Numerics" Version="4.11.0" />
|
||||
<PackageReference Include="TabletDriverPlugin" Version="0.3.0" />
|
||||
</ItemGroup>
|
||||
|
||||
</Project>
|
||||
|
|
@ -5,6 +5,8 @@ VisualStudioVersion = 16.0.30309.148
|
|||
MinimumVisualStudioVersion = 15.0.26124.0
|
||||
Project("{9A19103F-16F7-4668-BE54-9A1E7A4F7556}") = "AdvancedAntiSmoothingFilter", "AdvancedAntiSmoothingFilter\AdvancedAntiSmoothingFilter.csproj", "{57B25DD4-4DBC-4A63-84D0-7C50E33D3185}"
|
||||
EndProject
|
||||
Project("{9A19103F-16F7-4668-BE54-9A1E7A4F7556}") = "ExperimentalASFilter", "ExperimentalASFilter\ExperimentalASFilter.csproj", "{6421C523-5A5A-4F52-BB6B-A35F2381AA6E}"
|
||||
EndProject
|
||||
Global
|
||||
GlobalSection(SolutionConfigurationPlatforms) = preSolution
|
||||
Debug|Any CPU = Debug|Any CPU
|
||||
|
|
@ -15,6 +17,10 @@ Global
|
|||
{57B25DD4-4DBC-4A63-84D0-7C50E33D3185}.Debug|Any CPU.Build.0 = Debug|Any CPU
|
||||
{57B25DD4-4DBC-4A63-84D0-7C50E33D3185}.Release|Any CPU.ActiveCfg = Release|Any CPU
|
||||
{57B25DD4-4DBC-4A63-84D0-7C50E33D3185}.Release|Any CPU.Build.0 = Release|Any CPU
|
||||
{6421C523-5A5A-4F52-BB6B-A35F2381AA6E}.Debug|Any CPU.ActiveCfg = Debug|Any CPU
|
||||
{6421C523-5A5A-4F52-BB6B-A35F2381AA6E}.Debug|Any CPU.Build.0 = Debug|Any CPU
|
||||
{6421C523-5A5A-4F52-BB6B-A35F2381AA6E}.Release|Any CPU.ActiveCfg = Release|Any CPU
|
||||
{6421C523-5A5A-4F52-BB6B-A35F2381AA6E}.Release|Any CPU.Build.0 = Release|Any CPU
|
||||
EndGlobalSection
|
||||
GlobalSection(SolutionProperties) = preSolution
|
||||
HideSolutionNode = FALSE
|
||||
|
|
|
|||
Loading…
Add table
Reference in a new issue