wrong result

[rastinrastini]

Hi
i use code below to recognize but get wrong results for persian lang.
it must : ساغر می دی کفم نه تا ز بر
but recognized to:
Hypothesis: است
Hypothesis: يک شمه

Configuration configuration = new Configuration();

        configuration
                .setAcousticModelPath("file:/media/rastinrastini/3090AD6429530FCB/projects/speech/sphinx/hafez/model_parameters/hafez.ci_cont");
        configuration
                .setDictionaryPath("file:/media/rastinrastini/3090AD6429530FCB/projects/speech/sphinx/hafez/etc/hafez.dic");
        configuration
                .setLanguageModelPath("file:/media/rastinrastini/3090AD6429530FCB/projects/speech/sphinx/hafez/etc/hafez.lm");

        StreamSpeechRecognizer recognizer = new StreamSpeechRecognizer(
                configuration);
        InputStream stream = new FileInputStream(new File("/media/rastinrastini/3090AD6429530FCB/projects/speech/sphinx/hafez/wav/800.wav"));

        recognizer.startRecognition(stream);
        SpeechResult result;
        while ((result = recognizer.getResult()) != null) {
            System.out.format("Hypothesis: %s\n", result.getHypothesis());
        }
        recognizer.stopRecognition();

why it has this problem?
i have any wrong working?

[rastinrastini]

this is sound used for training and recognition.

[Nickolay V. Shmyrev]

One problem is the bandwidth of the signal which is less than required. You should have set the upperf during training properly

http://cmusphinx.sourceforge.net/wiki/faq/#qwhat_is_sample_rate_and_how_does_it_affect_accuracy

As for bad result, it might be caused by many other factors - language model quality, acoustic model quality, etc. You need to provide models and their source data in order to get help on this issue.