Shift Content description analysis to main analysis loop

Issue #252 Shift the calculation of content description from method SummarizeResults() to the actual Analysis loop. This is possible because the content description for each minute is independent of other minutes.

src/AnalysisConfigFiles/Towsey.TemplateDefinitions.json

@@ -213,7 +213,6 @@
       ]
     },
   },
-  {
   {
 	"Manifest": {
 	    "Name": "WindLight1",

src/AnalysisPrograms/AnalysisPrograms.csproj

@@ -302,7 +302,7 @@
     <Compile Include="..\AssemblyMetadata.cs">
       <Link>AssemblyMetadata.cs</Link>
     </Compile>
-    <Compile Include="ContentDescription.cs" />
+    <Compile Include="ContentDescription\ContentDescription.cs" />
     <Compile Include="AcousticWorkbench.Orchestration\EventMetadataResolver.cs" />
     <Compile Include="AcousticWorkbench.Orchestration\RemoteSegment.cs" />
     <Compile Include="AED.cs" />

src/AnalysisPrograms/ContentDescription.cs → src/AnalysisPrograms/ContentDescription/ContentDescription.cs

@@ -45,15 +45,33 @@ public class UseModel : AbstractStrongAnalyser
 
         public override string Identifier => TowseyContentDescription;
 
-        public override string Description => "[BETA] Generates six spectral indices for Content Description.";
+        public override string Description => "[BETA] Generates six spectral indices used as acoustic features to do Content Description.";
 
         public override AnalysisSettings DefaultSettings => new AnalysisSettings
         {
             AnalysisTargetSampleRate = ContentSignatures.SampleRate,
         };
 
+        private FunctionalTemplate[] functionalTemplates = null;
+
+        private Dictionary<string, Dictionary<string, double[]>> templatesAsDictionary = null;
+
         public override void BeforeAnalyze(AnalysisSettings analysisSettings)
         {
+            // Read in the functional templates file. These doe the content description.
+            var cdConfiguration = (CdConfig)analysisSettings.Configuration;
+            var cdConfigFile = analysisSettings.ConfigFile;
+            var configDirectory = cdConfigFile.DirectoryName ?? throw new ArgumentNullException(nameof(cdConfigFile), "Null value");
+            var templatesFileName = cdConfiguration.TemplatesList;
+            var templatesFile = new FileInfo(Path.Combine(configDirectory, templatesFileName));
+            this.functionalTemplates = Json.Deserialize<FunctionalTemplate[]>(templatesFile);
+            if (this.functionalTemplates == null)
+            {
+                throw new NullReferenceException(message: $"Array of functional templates was not read correctly from file: <{templatesFile}>");
+            }
+
+            // extract the template definitions as a dictionary. Each definition is itself a dictionary.
+            this.templatesAsDictionary = DataProcessing.ExtractDictionaryOfTemplateDictionaries(this.functionalTemplates);
         }
 
         public override AnalyzerConfig ParseConfig(FileInfo file) => ConfigFile.Deserialize<CdConfig>(file);
@@ -70,10 +88,20 @@ public class CdConfig : AnalyzerConfig
         }
 
         /// <summary>
-        /// This method calls IndexCalculateSixOnly.Analysis() to do the work!.
+        /// This method calls IndexCalculateSixOnly.Analysis() to calculate six spectral indices
+        /// and then calls ContentSignatures.AnalyzeOneMinute() to obtain a content description derived from those indices and an array of functional templates.
         /// </summary>
         public override AnalysisResult2 Analyze<T>(AnalysisSettings analysisSettings, SegmentSettings<T> segmentSettings)
         {
+            // set the start time for the current recording segment. Default is zero.
+            var elapsedTimeAtStartOfRecording = TimeSpan.Zero;
+            if (segmentSettings.SegmentStartOffset != null)
+            {
+                elapsedTimeAtStartOfRecording = (TimeSpan)segmentSettings.SegmentStartOffset;
+            }
+
+            var startMinuteId = (int)Math.Round(elapsedTimeAtStartOfRecording.TotalMinutes);
+
             var audioFile = segmentSettings.SegmentAudioFile;
             var recording = new AudioRecording(audioFile.FullName);
 
@@ -85,9 +113,25 @@ public override AnalysisResult2 Analyze<T>(AnalysisSettings analysisSettings, Se
 
             segmentResults.SpectralIndexValues.FileName = segmentSettings.Segment.SourceMetadata.Identifier;
 
-            // TODO: calculate description results here
-            DescriptionResult description = null; // e.g. GetContentDescription();
+            // DO THE CONTENT DESCRIPTION FOR ONE MINUTE HERE
+            // First get acoustic indices for one minute, convert to Dictionary and normalise the values.
+            var indicesDictionary = IndexCalculateSixOnly.ConvertIndicesToDictionary(segmentResults.SpectralIndexValues);
+            foreach (string key in ContentSignatures.IndexNames)
+            {
+                var indexBounds = ContentSignatures.IndexValueBounds[key];
+                var indexArray = indicesDictionary[key];
+                var normalisedVector = DataTools.NormaliseInZeroOne(indexArray, indexBounds[0], indexBounds[1]);
+                indicesDictionary[key] = normalisedVector;
+            }
 
+            // scan templates over one minute of indices to get content description
+            var descriptionResultForOneMinute = ContentSignatures.AnalyzeOneMinute(
+                this.functionalTemplates,
+                this.templatesAsDictionary,
+                indicesDictionary,
+                startMinuteId);
+
+            // set up the analysis results to return
             var analysisResults = new AnalysisResult2(analysisSettings, segmentSettings, recording.Duration)
             {
                 AnalysisIdentifier = this.Identifier,
@@ -99,7 +143,7 @@ public override AnalysisResult2 Analyze<T>(AnalysisSettings analysisSettings, Se
                 },
                 MiscellaneousResults =
                 {
-                    { nameof(DescriptionResult), description },
+                    { nameof(DescriptionResult), descriptionResultForOneMinute },
                 },
             };
 
@@ -164,13 +208,6 @@ public override void SummariseResults(
                 throw new NullReferenceException();
             }
 
-            // set the start time for the current recording segment. Default is zero.
-            var startTimeOffset = TimeSpan.Zero;
-            if (inputFileSegment.SegmentStartOffset != null)
-            {
-                startTimeOffset = (TimeSpan)inputFileSegment.SegmentStartOffset;
-            }
-
             // output config data to disk so other analyzers can use the data,
             // Should contain data only - i.e. the configuration settings that generated these indices
             // this data can then be used by later analysis processes.
@@ -199,9 +236,6 @@ public override void SummariseResults(
             Json.Serialise(icdPath.ToFileInfo(), indexConfigData);
 
             // gather spectra to form spectrograms.  Assume same spectra in all analyzer results
-            // this is the most efficient way to do this
-            // gather up numbers and strings store in memory, write to disk one time
-            // this method also AUTOMATICALLY SORTS because it uses array indexing
             var dictionaryOfSpectra = spectralIndices.ToTwoDimensionalArray(SpectralIndexValues.CachedSelectors, TwoDimensionalArray.Rotate90ClockWise);
 
             // Calculate the index distribution statistics and write to a json file. Also save as png image
@@ -217,19 +251,16 @@ public override void SummariseResults(
                     basename: basename,
                     indexSpectrograms: dictionaryOfSpectra);
 
-            // now get the content description for each minute.
-            var templatesFileName = cdConfiguration.TemplatesList;
-            var templatesFile = new FileInfo(Path.Combine(configDirectory, templatesFileName));
-
-            var contentDictionary = GetContentDescription(spectralIndices, templatesFile, startTimeOffset);
-            // TODO: use this when calculating content description results in the Analyze method
-            //var allContentDescriptionResults = results.Select(x => (DescriptionResult)x.MiscellaneousResults[nameof(DescriptionResult)]);
+            // Gather the content description results into an array of DescriptionResult and then convert to dictionary
+            var allContentDescriptionResults = results.Select(x => (DescriptionResult)x.MiscellaneousResults[nameof(DescriptionResult)]);
+            var contentDictionary = DataProcessing.ConvertResultsToDictionaryOfArrays(allContentDescriptionResults.ToList());
 
             // Write the results to a csv file
             var filePath = Path.Combine(resultsDirectory.FullName, "AcousticSignatures.csv");
             FileTools.WriteDictionaryAsCsvFile(contentDictionary, filePath);
-            // TODO: use this when calculating content description results in the Analyze method
-            //Csv.WriteToCsv(filePath, allContentDescriptionResults);
+
+            // TODO: fix this so it writes header and a column of content description values.
+            //Csv.WriteToCsv(new FileInfo(filePath), contentDictionary);
 
             // prepare graphical plots of the acoustic signatures.
             var contentPlots = GetPlots(contentDictionary);
@@ -245,6 +276,7 @@ public override void SummariseResults(
         }
 
         /// <summary>
+        /// NOTE: THIS METHOD SHOULD EVENTUALLY BE DELETED. NO LONGER CALLED.
         /// Calculate the content description for each minute.
         /// </summary>
         /// <param name="spectralIndices">set of spectral indices for each minute.</param>

src/AudioAnalysisTools/AudioAnalysisTools.csproj

@@ -246,6 +246,7 @@
     <Compile Include="ContentDescriptionTools\ContentVisualization.cs" />
     <Compile Include="ContentDescriptionTools\DataProcessing.cs" />
     <Compile Include="ContentDescriptionTools\DescriptionResult.cs" />
+    <Compile Include="ContentDescriptionTools\SpectralIndexValuesforContentDescription.cs" />
     <Compile Include="ContentDescriptionTools\TemplateManifest.cs" />
     <Compile Include="ContentDescriptionTools\TemplateCollection.cs" />
     <Compile Include="CrossCorrelation.cs" />

src/AudioAnalysisTools/ContentDescriptionTools/DataProcessing.cs

@@ -361,12 +361,19 @@ public static double[] GetFreqBinVector(Dictionary<string, double[]> dictionary,
             return list.ToArray();
         }
 
+        public static Dictionary<string, double[]> ConvertResultsToDictionaryOfArrays(List<DescriptionResult> results)
+        {
+            int arrayLength = results.Count;
+            int arrayStart = 0;
+            return ConvertResultsToDictionaryOfArrays(results, arrayLength, arrayStart);
+        }
+
         /// <summary>
-        /// Converts individual results to a dictionary of plots.
-        /// </summary>
-        /// <param name="results">a list of results for each content type in every minute.</param>
-        /// <param name="arrayLength">The plot length will the total number of minutes scanned, typically 1440 for one day.</param>
-        /// <param name="arrayStart">time start.</param>
+         /// Converts individual results to a dictionary of plots.
+         /// </summary>
+         /// <param name="results">a list of results for each content type in every minute.</param>
+         /// <param name="arrayLength">The plot length will the total number of minutes scanned, typically 1440 for one day.</param>
+         /// <param name="arrayStart">time start.</param>
         public static Dictionary<string, double[]> ConvertResultsToDictionaryOfArrays(List<DescriptionResult> results, int arrayLength, int arrayStart)
         {
             var arrays = new Dictionary<string, double[]>();

src/AudioAnalysisTools/ContentDescriptionTools/TemplateManifest.cs

@@ -25,7 +25,7 @@ namespace AudioAnalysisTools.ContentDescriptionTools
     public enum EditStatus
     {
         Edit,   // This option edits an existing functional template in the json file. The template definition is (re)calculated.
-        Copy,   // This option keeps an existing functional template unchanged except for its template Id. Template Ids are assigned in order they appear in the yml file.
+        Copy,   // This option keeps an existing functional template unchanged.
         Ignore, // This option keeps an existing functional template unchanged except changes its UseStatus boolean field to FALSE.
     }
 
@@ -60,7 +60,6 @@ public static void CreateNewFileOfTemplateDefinitions(FileInfo manifestFile, Fil
                 {
                     // the current manifest is not an existing template - therefore make it.
                     var newTemplate = CreateNewTemplateFromManifest(manifest);
-                    newTemplate.TemplateId = i;
                     newTemplate.Template = CreateTemplateDefinition(manifest);
                     newTemplate.MostRecentEdit = DateTime.Now;
                     newTemplateList.Add(newTemplate);
@@ -71,7 +70,6 @@ public static void CreateNewFileOfTemplateDefinitions(FileInfo manifestFile, Fil
                 {
                     // This option edits an existing functional template in the json file. The template definition is (re)calculated.
                     var newTemplate = CreateNewTemplateFromManifest(manifest);
-                    newTemplate.TemplateId = i;
                     newTemplate.Template = CreateTemplateDefinition(manifest);
                     newTemplate.MostRecentEdit = DateTime.Now;
                     newTemplateList.Add(newTemplate);
@@ -82,10 +80,8 @@ public static void CreateNewFileOfTemplateDefinitions(FileInfo manifestFile, Fil
                 {
                     // TODO: intentionally broken. FunctionalTemplates should be immutable. If they need to change create a new one (could be a copy, but it would have a version or edit date etc...).
                     throw new NotImplementedException();
-                    // This option keeps an existing functional template unchanged except for its template Id.
-                    // Template ids are assigned in order they appear in the yml file.
+                    // This option keeps an existing functional template unchanged.
                     //var existingTemplate = dictionaryOfCurrentTemplates[name];
-                    //existingTemplate.TemplateId = i;
                     //existingTemplate.UseStatus = true;
                     //existingTemplate.Provenance = null;
                     //newTemplateList.Add(existingTemplate);
@@ -99,7 +95,6 @@ public static void CreateNewFileOfTemplateDefinitions(FileInfo manifestFile, Fil
                     throw new NotImplementedException();
                     // This option keeps an existing functional template unchanged except changes its UseStatus boolean field to FALSE.
                     //var existingTemplate = dictionaryOfCurrentTemplates[name];
-                    //existingTemplate.TemplateId = i;
                     //existingTemplate.Provenance = null;
                     //existingTemplate.UseStatus = false;
                     //newTemplateList.Add(existingTemplate);
@@ -281,9 +276,6 @@ public class FunctionalTemplate
     {
         public TemplateManifest Manifest { get; set; }
 
-        // TODO: I'd suggest removing this concept. Integer based IDs are brittle.
-        public int TemplateId { get; set; }
-
         /// <summary>
         /// Gets or sets a value indicating whether to use the template or not.
         /// UseStatus can be true or false.