Azure recognizer language detection added

PaciStardust · Sep 30, 2022 · 9efdae8 · 9efdae8
1 parent 63ac1f1
commit 9efdae8
Show file tree

Hide file tree

Showing 6 changed files with 42 additions and 20 deletions.
diff --git a/OscMultitool/Config.cs b/OscMultitool/Config.cs
@@ -288,10 +288,11 @@ public int TranslationMaxTextLength
             //Azure
             public string AzureRegion { get; set; } = string.Empty;
             public string AzureKey { get; set; } = string.Empty;
-            public string AzureLanguage { get; set; } = string.Empty;
+            public string AzureSpeechLanguage { get; set; } = string.Empty;
             public string AzureCustomEndpoint { get; set; } = string.Empty;
             public string AzureVoice { get; set; } = string.Empty;
             public List<string> AzurePhrases { get; set; } = new();
+            public List<string> AzureRecognitionLanguages { get; set; } = new();
 
             //Usage
             public bool TranslateTts { get; set; } = false;

diff --git a/OscMultitool/Services/Api/Synthesizer.cs b/OscMultitool/Services/Api/Synthesizer.cs
@@ -44,8 +44,8 @@ public static void ReloadClient()
                 if (!string.IsNullOrWhiteSpace(Config.Api.AzureVoice))
                     speechCfg.SpeechSynthesisVoiceName = Config.Api.AzureVoice;
 
-                if (!string.IsNullOrWhiteSpace(Config.Api.AzureLanguage))
-                    speechCfg.SpeechSynthesisLanguage = Config.Api.AzureLanguage;
+                if (!string.IsNullOrWhiteSpace(Config.Api.AzureSpeechLanguage))
+                    speechCfg.SpeechSynthesisLanguage = Config.Api.AzureSpeechLanguage;
 
                 speechCfg.SetSpeechSynthesisOutputFormat(SpeechSynthesisOutputFormat.Raw16Khz16BitMonoPcm);
 

diff --git a/OscMultitool/Services/Speech/Recognizers/RecognizerAzure.cs b/OscMultitool/Services/Speech/Recognizers/RecognizerAzure.cs
@@ -43,14 +43,25 @@ protected override bool StartInternal()
                 if (!string.IsNullOrWhiteSpace(Config.Api.AzureCustomEndpoint))
                     speechConfig.EndpointId = Config.Api.AzureCustomEndpoint;
 
-                if (!string.IsNullOrWhiteSpace(Config.Api.AzureLanguage))
-                    speechConfig.SpeechRecognitionLanguage = Config.Api.AzureLanguage;
-
-                rec = new(speechConfig, audioConfig);
+                if (Config.Api.AzureRecognitionLanguages.Count > 1)
+                {
+                    speechConfig.SetProperty(PropertyId.SpeechServiceConnection_ContinuousLanguageIdPriority, "Latency");
+                    var autoDetectSourceLanguageConfig = AutoDetectSourceLanguageConfig.FromLanguages(Config.Api.AzureRecognitionLanguages.ToArray());
+                    rec = new(speechConfig, autoDetectSourceLanguageConfig, audioConfig);
+                }
+                else
+                {
+                    if (Config.Api.AzureRecognitionLanguages.Count == 1)
+                        speechConfig.SpeechRecognitionLanguage = Config.Api.AzureRecognitionLanguages[0];
+                    rec = new(speechConfig, audioConfig);
+                }      
 
-                var phraseList = PhraseListGrammar.FromRecognizer(rec);
-                foreach(var phrase in Config.Api.AzurePhrases)
-                    phraseList.AddPhrase(phrase);
+                if (Config.Api.AzurePhrases.Count != 0)
+                {
+                    var phraseList = PhraseListGrammar.FromRecognizer(rec);
+                    foreach (var phrase in Config.Api.AzurePhrases)
+                        phraseList.AddPhrase(phrase);
+                }
             }
             catch (Exception e)
             {
@@ -123,9 +134,13 @@ private async Task StartRecognizing()
         #region Events
         private void OnRecognized(object? sender, SpeechRecognitionEventArgs e)
         {
-            Logger.Log("Got Message: " + e.Result.Text);
+            var result = e.Result.Text;
+            if (string.IsNullOrWhiteSpace(result))
+                return;
+
+            Logger.Log("Got Message: " + result);
 
-            var message = Denoise(e.Result.Text);
+            var message = Denoise(result);
             if (string.IsNullOrWhiteSpace(message))
                 return;
 

diff --git a/OscMultitool/Ui/Pages/PageApi.xaml b/OscMultitool/Ui/Pages/PageApi.xaml
@@ -91,21 +91,24 @@
                     <TextBlock Style="{DynamicResource HTextBlock}" Text="Custom endpoint (Optional)" Margin="0,4,0,0"/>
                     <TextBox Text="{Binding Api.AzureCustomEndpoint, Source={x:Static local:Config.Data}}" Tag="Custom endpoint..." Style="{StaticResource HTextBoxLong}" PreviewKeyDown="TextBox_PreviewKeyDown"/>
 
-                    <WrapPanel Margin="0,4,0,0">
-                        <TextBox HorizontalAlignment="Left" Text="{Binding Api.AzureLanguage, Source={x:Static local:Config.Data}}" Tag="Language..." Style="{StaticResource HTextBoxShort}" PreviewKeyDown="TextBox_PreviewKeyDown"/>
-                        <Label Content="Used Language (Optional)" VerticalAlignment="Center"/>
-                    </WrapPanel>
-
                     <Label Content="Speech Recognition Options (Optional)" Margin="0,8,0,0"/>
                     <StackPanel Style="{StaticResource HStackIndent}">
-                        <Button Content="Edit phrase list" Click="Button_EditPhrases" HorizontalAlignment="Left"/>
+                        <WrapPanel>
+                            <Button Content="Edit phrases" Click="Button_EditPhrases"/>
+                            <Button Content="Edit languages" Click="Button_EditLanguages" Margin="16,0,0,0"/>
+                        </WrapPanel>
                     </StackPanel>
 
                     <Label Content="Text-to-Speech Options (Optional)" Margin="0,8,0,0"/>
                     <StackPanel Style="{StaticResource HStackIndent}">
                         <CheckBox Content="Use Azure TTS" IsChecked="{Binding Api.UseAzureTts, Source={x:Static local:Config.Data}}"/>
                         <TextBlock Style="{DynamicResource HTextBlock}" Text="Output Speech Voice" Margin="0,4,0,0"/>
                         <TextBox Text="{Binding Api.AzureVoice, Source={x:Static local:Config.Data}}" Tag="Speech Voice..." Style="{StaticResource HTextBoxLong}" PreviewKeyDown="TextBox_PreviewKeyDown"/>
+
+                        <WrapPanel Margin="0,4,0,0">
+                            <TextBox HorizontalAlignment="Left" Text="{Binding Api.AzureSpeechLanguage, Source={x:Static local:Config.Data}}" Tag="Language..." Style="{StaticResource HTextBoxShort}" PreviewKeyDown="TextBox_PreviewKeyDown"/>
+                            <Label Content="Used Language" VerticalAlignment="Center"/>
+                        </WrapPanel>
                     </StackPanel>
                 </StackPanel>
             </StackPanel>

diff --git a/OscMultitool/Ui/Pages/PageApi.xaml.cs b/OscMultitool/Ui/Pages/PageApi.xaml.cs
@@ -71,6 +71,9 @@ private void Button_ReloadSynthesizer(object sender, RoutedEventArgs e)
 
         private void Button_EditPhrases(object sender, RoutedEventArgs e)
             => UiHelper.OpenListEditor("Edit phrases", "Phrase", Config.Api.AzurePhrases, "New Phrase");
+
+        private void Button_EditLanguages(object sender, RoutedEventArgs e)
+            => UiHelper.OpenListEditor("Edit languages", "Language", Config.Api.AzureRecognitionLanguages, "New Language");
         #endregion
 
         #region SelectionChanged

diff --git a/OscMultitool/Ui/Pages/PageConfig.xaml b/OscMultitool/Ui/Pages/PageConfig.xaml
@@ -40,8 +40,8 @@
                     <CheckBox Content="Automatically check for updates" IsChecked="{Binding Debug.CheckUpdates, Source={x:Static local:Config.Data}}" Margin="0,4,0,0"/>
                     <WrapPanel Margin="0,4,0,0">
                         <Button Content="Open Docs" HorizontalAlignment="Left" Click="Button_OpenDocs"/>
-                        <Button Margin="8,0,0,0" Content="Open Config" HorizontalAlignment="Left" Click="Button_OpenConfig"/>
-                        <Button Margin="8,0,0,0" Content="Check Update" HorizontalAlignment="Left" Click="Button_CheckUpdate"/>
+                        <Button Margin="16,0,0,0" Content="Open Config" HorizontalAlignment="Left" Click="Button_OpenConfig"/>
+                        <Button Margin="16,0,0,0" Content="Check Update" HorizontalAlignment="Left" Click="Button_CheckUpdate"/>
                     </WrapPanel>
                     <Button Margin="0,8,0,0" Content="Reload devices" Foreground="{x:Static ui:UiHelper.ColorInvalid}" HorizontalAlignment="Left" Click="Button_ReloadDevices"/>
                 </StackPanel>