Explore Courses Blog Tutorials Interview Questions
0 votes
in Python by (12.7k points)

I am chipping away at a task that includes speech recognition utilizing the SpeechRecognition module. One thing I need to do to improve my speech recognition is to have the option to yield the words that have been perceived at the earliest opportunity. I need it to be like at whatever point you talk into Google Translate when you state a word it yields it on the screen to tell you that you have said it. 

A portion of the things I have attempted is to have an exhibit that stores separate sound chronicles and have speech recognition emphasize through the cluster perceiving every sound account and afterward yielding that. This didn't work on the grounds that various words take various occasions to state. 

I looked further into the Google API for speech recognition given to me by the SpeechRecognition module and needed to perceive how I could change the genuine library by adding print articulations in certain spots to accomplish the objective. I didn't have the foggiest idea where to put, as I am a novice in speech recognition and that I don't think a lot about the Google Speech Recognition API. 

Here is the google programming interface code, it gets to the cloud to do sr.

def recognize_google(self, audio_data, key=None, language="en-US", show_all=False):


        Performs speech recognition on ``audio_data`` (an ``AudioData`` instance), using the Google Speech Recognition API.

        The Google Speech Recognition API key is specified by ``key``. If not specified, it uses a generic key that works out of the box. This should generally be used for personal or testing purposes only, as it **may be revoked by Google at any time**.

        To obtain your own API key, simply following the steps on the `API Keys <>`__ page at the Chromium Developers site. In the Google Developers Console, Google Speech Recognition is listed as "Speech API".

        The recognition language is determined by ``language``, an RFC5646 language tag like ``"en-US"`` (US English) or ``"fr-FR"`` (International French), defaulting to US English. A list of supported language tags can be found in this `StackOverflow answer <>`__.

        Returns the most likely transcription if ``show_all`` is false (the default). Otherwise, returns the raw API response as a JSON dictionary.

        Raises a ``speech_recognition.UnknownValueError`` exception if the speech is unintelligible. Raises a ``speech_recognition.RequestError`` exception if the speech recognition operation failed, if the key isn't valid, or if there is no internet connection.


        assert isinstance(audio_data, AudioData), "``audio_data`` must be audio data"

        assert key is None or isinstance(key, str), "``key`` must be ``None`` or a string"

        assert isinstance(language, str), "``language`` must be a string"

        flac_data = audio_data.get_flac_data(

            convert_rate=None if audio_data.sample_rate >= 8000 else 8000,  # audio samples must be at least 8 kHz

            convert_width=2  # audio samples must be 16-bit


        if key is None: key = "AIzaSyBOti4mM-6x9WDnZIjIeyEU21OpBXqWBgw"

        url = "{}".format(urlencode({

            "client": "chromium",

            "lang": language,

            "key": key,


        request = Request(url, data=flac_data, headers={"Content-Type": "audio/x-flac; rate={}".format(audio_data.sample_rate)})

        # obtain audio transcription results


            response = urlopen(request, timeout=self.operation_timeout)

        except HTTPError as e:

            raise RequestError("recognition request failed: {}".format(e.reason))

        except URLError as e:

            raise RequestError("recognition connection failed: {}".format(e.reason))

        response_text ="utf-8")

        # ignore any blank blocks

        actual_result = []

        for line in response_text.split("\n"):

            if not line: continue

            result = json.loads(line)["result"]

            if len(result) != 0:

                actual_result = result[0]




        # return results

        if show_all: return actual_result

        if not isinstance(actual_result, dict) or len(actual_result.get("alternative", [])) == 0: raise UnknownValueError()

        if "confidence" in actual_result["alternative"]:

            # return alternative with highest confidence score

            best_hypothesis = max(actual_result["alternative"], key=lambda alternative: alternative["confidence"])


            # when there is no confidence available, we arbitrarily choose the first hypothesis.

            best_hypothesis = actual_result["alternative"][0]

        if "transcript" not in best_hypothesis: raise UnknownValueError()

        return best_hypothesis["transcript"]

Look at my base code, It can do speech recognition successfully.
r = sr.Recognizer()
m = sr.Microphone(); 
r = sr.Recognizer()
on = True 
while on :
    with sr.Microphone() as source:
        audio = r.listen(source)
            text = r.recognize_google(audio)
            print("You said: {}".format(text))
            print("Sorry, we did not recognize your voice")
This is the final method, that's recording function to make audio files or objects:
def listen(self, source, timeout=None, phrase_time_limit=None, snowboy_configuration=None):
        Records a single phrase from ``source`` (an ``AudioSource`` instance) into an ``AudioData`` instance, which it returns.
        This is done by waiting until the audio has an energy above ``recognizer_instance.energy_threshold`` (the user has started speaking), and then recording until it encounters ``recognizer_instance.pause_threshold`` seconds of non-speaking or there is no more audio input. The ending silence is not included.
        The ``timeout`` parameter is the maximum number of seconds that this will wait for a phrase to start before giving up and throwing an ``speech_recognition.WaitTimeoutError`` exception. If ``timeout`` is ``None``, there will be no wait timeout.
        The ``phrase_time_limit`` parameter is the maximum number of seconds that this will allow a phrase to continue before stopping and returning the part of the phrase processed before the time limit was reached. The resulting audio will be the phrase cut off at the time limit. If ``phrase_timeout`` is ``None``, there will be no phrase time limit.
        The ``snowboy_configuration`` parameter allows integration with `Snowboy <>`__, an offline, high-accuracy, power-efficient hotword recognition engine. When used, this function will pause until Snowboy detects a hotword, after which it will unpause. This parameter should either be ``None`` to turn off Snowboy support, or a tuple of the form ``(SNOWBOY_LOCATION, LIST_OF_HOT_WORD_FILES)``, where ``SNOWBOY_LOCATION`` is the path to the Snowboy root directory, and ``LIST_OF_HOT_WORD_FILES`` is a list of paths to Snowboy hotword configuration files (`*.pmdl` or `*.umdl` format).
        This operation will always complete within ``timeout + phrase_timeout`` seconds if both are numbers, either by returning the audio data, or by raising a ``speech_recognition.WaitTimeoutError`` exception.
        assert isinstance(source, AudioSource), "Source must be an audio source"
        assert is not None, "Audio source must be entered before listening, see documentation for ``AudioSource``; are you using ``source`` outside of a ``with`` statement?"
        assert self.pause_threshold >= self.non_speaking_duration >= 0
        if snowboy_configuration is not None:
            assert os.path.isfile(os.path.join(snowboy_configuration[0], "")), "``snowboy_configuration[0]`` must be a Snowboy root directory containing ````"
            for hot_word_file in snowboy_configuration[1]:
                assert os.path.isfile(hot_word_file), "``snowboy_configuration[1]`` must be a list of Snowboy hot word configuration files"
        seconds_per_buffer = float(source.CHUNK) / source.SAMPLE_RATE
        pause_buffer_count = int(math.ceil(self.pause_threshold / seconds_per_buffer))  # number of buffers of non-speaking audio during a phrase, before the phrase should be considered complete
        phrase_buffer_count = int(math.ceil(self.phrase_threshold / seconds_per_buffer))  # minimum number of buffers of speaking audio before we consider the speaking audio a phrase
        non_speaking_buffer_count = int(math.ceil(self.non_speaking_duration / seconds_per_buffer))  # maximum number of buffers of non-speaking audio to retain before and after a phrase
        # read audio input for phrases until there is a phrase that is long enough
        elapsed_time = 0  # number of seconds of audio read
        buffer = b""  # an empty buffer means that the stream has ended and there is no data left to read
        while True:
            frames = collections.deque()
            if snowboy_configuration is None:
                # store audio input until the phrase starts
                while True:
                    # handle waiting too long for phrase by raising an exception
                    elapsed_time += seconds_per_buffer
                    if timeout and elapsed_time > timeout:
                        raise WaitTimeoutError("listening timed out while waiting for phrase to start")
                    buffer =
                    if len(buffer) == 0: break  # reached end of the stream
                    if len(frames) > non_speaking_buffer_count:  # ensure we only keep the needed amount of non-speaking buffers
                    # detect whether speaking has started on audio input
                    energy = audioop.rms(buffer, source.SAMPLE_WIDTH)  # energy of the audio signal
                    if energy > self.energy_threshold: break
                    # dynamically adjust the energy threshold using asymmetric weighted average
                    if self.dynamic_energy_threshold:
                        damping = self.dynamic_energy_adjustment_damping ** seconds_per_buffer  # account for different chunk sizes and rates
                        target_energy = energy * self.dynamic_energy_ratio
                        self.energy_threshold = self.energy_threshold * damping + target_energy * (1 - damping)
                # read audio input until the hotword is said
                snowboy_location, snowboy_hot_word_files = snowboy_configuration
                buffer, delta_time = self.snowboy_wait_for_hot_word(snowboy_location, snowboy_hot_word_files, source, timeout)
                elapsed_time += delta_time
                if len(buffer) == 0: break  # reached end of the stream
            # read audio input until the phrase ends
            pause_count, phrase_count = 0, 0
            phrase_start_time = elapsed_time
            while True:
                # handle phrase being too long by cutting off the audio
                elapsed_time += seconds_per_buffer
                if phrase_time_limit and elapsed_time - phrase_start_time > phrase_time_limit:
                buffer =
                if len(buffer) == 0: break  # reached end of the stream
                phrase_count += 1
                # check if speaking has stopped for longer than the pause threshold on the audio input
                energy = audioop.rms(buffer, source.SAMPLE_WIDTH)  # unit energy of the audio signal within the buffer
                if energy > self.energy_threshold:
                    pause_count = 0
                    pause_count += 1
                if pause_count > pause_buffer_count:  # end of the phrase
            # check how long the detected phrase is, and retry listening if the phrase is too short
            phrase_count -= pause_count  # exclude the buffers for the pause before the phrase
            if phrase_count >= phrase_buffer_count or len(buffer) == 0: break  # phrase is long enough or we've reached the end of the stream, so stop listening
        # obtain frame data
        for i in range(pause_count - non_speaking_buffer_count): frames.pop()  # remove extra non-speaking frames at the end
        frame_data = b"".join(frames)
        return AudioData(frame_data, source.SAMPLE_RATE, source.SAMPLE_WIDTH)
I want a software which takes the above code and implement in such a way so that it outputs the word which is recognized as soon as possible, Which is similar to Google translator
python speech-recognition speech-to-text

1 Answer

0 votes
by (26.4k points)

Have you seen this before ? Click on this link

There just click "record audio", you can able to see the hypotheses on the screen while you speak.

This demo is actually a Open source, you can just fork the code in GitHub. This is achieved with the help of WebSocket API using your favorite programming language.

Want to know more about Python? Come and learn: python course

Browse Categories