1

我正在通过我的应用程序中的 OpenEars 功能实现语音到文本。我还使用Rejecto插件来更好地识别并RapidEars获得更快的结果。目标是检测短语和单个单词，例如：

    lmGenerator = [[LanguageModelGenerator alloc] init];

NSArray *words = [NSArray arrayWithObjects:@"REBETANDEAL",@"NEWBET",@"REEEBET", nil];
NSString *name = @"NameIWantForMyLanguageModelFiles";
NSError *err = [lmGenerator generateRejectingLanguageModelFromArray:words
                                                     withFilesNamed:name
                                             withOptionalExclusions:nil
                                                    usingVowelsOnly:FALSE
                                                         withWeight:nil
                                             forAcousticModelAtPath:[AcousticModel pathToModel:@"AcousticModelEnglish"]]; // Change "AcousticModelEnglish" to "AcousticModelSpanish" to create a Spanish Rejecto model.
// Change "AcousticModelEnglish" to "AcousticModelSpanish" to create a Spanish language model instead of an English one.

NSDictionary *languageGeneratorResults = nil;

NSString *lmPath = nil;
NSString *dicPath = nil;

if([err code] == noErr) {

    languageGeneratorResults = [err userInfo];

    lmPath = [languageGeneratorResults objectForKey:@"LMPath"];
    dicPath = [languageGeneratorResults objectForKey:@"DictionaryPath"];

} else {
    NSLog(@"Error: %@",[err localizedDescription]);
}





// Change "AcousticModelEnglish" to "AcousticModelSpanish" to perform Spanish recognition instead of English.
[self.pocketsphinxController setRapidEarsToVerbose:FALSE]; // This defaults to FALSE but will give a lot of debug readout if set TRUE
[self.pocketsphinxController setRapidEarsAccuracy:10]; // This defaults to 20, maximum accuracy, but can be set as low as 1 to save CPU
[self.pocketsphinxController setFinalizeHypothesis:TRUE]; // This defaults to TRUE and will return a final hypothesis, but can be turned off to save a little CPU and will then return no final hypothesis; only partial "live" hypotheses.
[self.pocketsphinxController setFasterPartials:TRUE]; // This will give faster rapid recognition with less accuracy. This is what you want in most cases since more accuracy for partial hypotheses will have a delay.
[self.pocketsphinxController setFasterFinals:FALSE]; // This will give an accurate final recognition. You can have earlier final recognitions with less accuracy as well by setting this to TRUE.
[self.pocketsphinxController startRealtimeListeningWithLanguageModelAtPath:lmPath dictionaryAtPath:dicPath acousticModelAtPath:[AcousticModel pathToModel:@"AcousticModelEnglish"]]; // Starts the rapid recognition loop. Change "AcousticModelEnglish" to "AcousticModelSpanish" in order to perform Spanish language recognition.


[self.openEarsEventsObserver setDelegate:self];

大多数时候结果很好，但有时它会从单独的字符串对象中混合。例如我传递words数组：@[@"ME AND YOU",@"YOU",@"ME"]并且输出可以是："YOU ME ME ME AND"。我不希望它只识别短语的一部分。请问有什么想法吗？

4

2 回答 2

1

OpenEars 开发人员在这里。要使用 OpenEars 检测固定短语，请使用 LanguageModelGenerator 的新动态语法生成器方法动态创建基于规则的语法而不是统计语言模型：http ://www.politepix.com/2014/04/10/openears-1 -7-引入动态语法生成/

于 2014-04-28T18:31:13.097 回答
1

您可以在pocketsphinxDidReceiveHypothesis:(NSString *)hypothesis recognitionScore:(NSString *)recognitionScore utteranceID:(NSString *)utteranceID显示假设之前检查假设是否在您的单词数组中。

- (void) pocketsphinxDidReceiveHypothesis:(NSString *)hypothesis recognitionScore:(NSString *)recognitionScore utteranceID:(NSString *)utteranceID {
            if ([words containsObject:hypothesis]) {
                  //show hypothesis
            }           
}
于 2014-01-10T18:20:56.890 回答