Detectar y grabar audio en Python

Necesito capturar clips de audio como archivos WAV que luego puedo pasar a otro bit de Python para su procesamiento. El problema es que necesito determinar cuándo hay audio presente y luego grabarlo, detenerme cuando se quede en silencio y luego pasar ese archivo al módulo de procesamiento.

Estoy pensando que debería ser posible con el módulo de ondas detectar cuando hay silencio puro y descartarlo, luego, tan pronto como se detecte algo que no sea silencio, comience a grabar, luego, cuando la línea se vuelva silenciosa, detenga la grabación.

Simplemente no puedo entenderlo, ¿alguien puede ayudarme a comenzar con un ejemplo básico?

Como continuación de la respuesta de Nick Fortescue, aquí hay un ejemplo más completo de cómo grabar desde el micrófono y procesar los datos resultantes:

from sys import byteorder
from array import array
from struct import pack

import pyaudio
import wave

THRESHOLD = 500
CHUNK_SIZE = 1024
FORMAT = pyaudio.paInt16
RATE = 44100

def is_silent(snd_data):
    "Returns 'True' if below the 'silent' threshold"
    return max(snd_data) < THRESHOLD

def normalize(snd_data):
    "Average the volume out"
    MAXIMUM = 16384
    times = float(MAXIMUM)/max(abs(i) for i in snd_data)

    r = array('h')
    for i in snd_data:
        r.append(int(i*times))
    return r

def trim(snd_data):
    "Trim the blank spots at the start and end"
    def _trim(snd_data):
        snd_started = False
        r = array('h')

        for i in snd_data:
            if not snd_started and abs(i)>THRESHOLD:
                snd_started = True
                r.append(i)

            elif snd_started:
                r.append(i)
        return r

    # Trim to the left
    snd_data = _trim(snd_data)

    # Trim to the right
    snd_data.reverse()
    snd_data = _trim(snd_data)
    snd_data.reverse()
    return snd_data

def add_silence(snd_data, seconds):
    "Add silence to the start and end of 'snd_data' of length 'seconds' (float)"
    silence = [0] * int(seconds * RATE)
    r = array('h', silence)
    r.extend(snd_data)
    r.extend(silence)
    return r

def record():
    """
    Record a word or words from the microphone and 
    return the data as an array of signed shorts.

    Normalizes the audio, trims silence from the 
    start and end, and pads with 0.5 seconds of 
    blank sound to make sure VLC et al can play 
    it without getting chopped off.
    """
    p = pyaudio.PyAudio()
    stream = p.open(format=FORMAT, channels=1, rate=RATE,
        input=True, output=True,
        frames_per_buffer=CHUNK_SIZE)

    num_silent = 0
    snd_started = False

    r = array('h')

    while 1:
        # little endian, signed short
        snd_data = array('h', stream.read(CHUNK_SIZE))
        if byteorder == 'big':
            snd_data.byteswap()
        r.extend(snd_data)

        silent = is_silent(snd_data)

        if silent and snd_started:
            num_silent += 1
        elif not silent and not snd_started:
            snd_started = True

        if snd_started and num_silent > 30:
            break

    sample_width = p.get_sample_size(FORMAT)
    stream.stop_stream()
    stream.close()
    p.terminate()

    r = normalize(r)
    r = trim(r)
    r = add_silence(r, 0.5)
    return sample_width, r

def record_to_file(path):
    "Records from the microphone and outputs the resulting data to 'path'"
    sample_width, data = record()
    data = pack('<' + ('h'*len(data)), *data)

    wf = wave.open(path, 'wb')
    wf.setnchannels(1)
    wf.setsampwidth(sample_width)
    wf.setframerate(RATE)
    wf.writeframes(data)
    wf.close()

if __name__ == '__main__':
    print("please speak a word into the microphone")
    record_to_file('demo.wav')
    print("done - result written to demo.wav")

Creo que el módulo WAVE no admite la grabación, solo procesa archivos existentes. Es posible que desee mirar PyAudio para grabar realmente. WAV es el formato de archivo más simple del mundo. En paInt16 solo obtienes un entero con signo que representa un nivel, y más cerca de 0 es más silencioso. No recuerdo si los archivos WAV tienen un byte alto primero o un byte bajo, pero algo como esto debería funcionar (lo siento, no soy realmente un programador de Python:

from array import array

# you'll probably want to experiment on threshold
# depends how noisy the signal
threshold = 10 
max_value = 0

as_ints = array('h', data)
max_value = max(as_ints)
if max_value > threshold:
    # not silence

El código PyAudio para la grabación se mantiene como referencia:

import pyaudio
import sys

chunk = 1024
FORMAT = pyaudio.paInt16
CHANNELS = 1
RATE = 44100
RECORD_SECONDS = 5

p = pyaudio.PyAudio()

stream = p.open(format=FORMAT,
                channels=CHANNELS, 
                rate=RATE, 
                input=True,
                output=True,
                frames_per_buffer=chunk)

print "* recording"
for i in range(0, 44100 / chunk * RECORD_SECONDS):
    data = stream.read(chunk)
    # check for silence here by comparing the level with 0 (or some threshold) for 
    # the contents of data.
    # then write data or not to a file

print "* done"

stream.stop_stream()
stream.close()
p.terminate()

Gracias a cryo por la versión mejorada que basé mi código probado a continuación:

#Instead of adding silence at start and end of recording (values=0) I add the original audio . This makes audio sound more natural as volume is >0. See trim()
#I also fixed issue with the previous code - accumulated silence counter needs to be cleared once recording is resumed.

from array import array
from struct import pack
from sys import byteorder
import copy
import pyaudio
import wave

THRESHOLD = 500  # audio levels not normalised.
CHUNK_SIZE = 1024
SILENT_CHUNKS = 3 * 44100 / 1024  # about 3sec
FORMAT = pyaudio.paInt16
FRAME_MAX_VALUE = 2 ** 15 - 1
NORMALIZE_MINUS_ONE_dB = 10 ** (-1.0 / 20)
RATE = 44100
CHANNELS = 1
TRIM_APPEND = RATE / 4

def is_silent(data_chunk):
    """Returns 'True' if below the 'silent' threshold"""
    return max(data_chunk) < THRESHOLD

def normalize(data_all):
    """Amplify the volume out to max -1dB"""
    # MAXIMUM = 16384
    normalize_factor = (float(NORMALIZE_MINUS_ONE_dB * FRAME_MAX_VALUE)
                        / max(abs(i) for i in data_all))

    r = array('h')
    for i in data_all:
        r.append(int(i * normalize_factor))
    return r

def trim(data_all):
    _from = 0
    _to = len(data_all) - 1
    for i, b in enumerate(data_all):
        if abs(b) > THRESHOLD:
            _from = max(0, i - TRIM_APPEND)
            break

    for i, b in enumerate(reversed(data_all)):
        if abs(b) > THRESHOLD:
            _to = min(len(data_all) - 1, len(data_all) - 1 - i + TRIM_APPEND)
            break

    return copy.deepcopy(data_all[_from:(_to + 1)])

def record():
    """Record a word or words from the microphone and 
    return the data as an array of signed shorts."""

    p = pyaudio.PyAudio()
    stream = p.open(format=FORMAT, channels=CHANNELS, rate=RATE, input=True, output=True, frames_per_buffer=CHUNK_SIZE)

    silent_chunks = 0
    audio_started = False
    data_all = array('h')

    while True:
        # little endian, signed short
        data_chunk = array('h', stream.read(CHUNK_SIZE))
        if byteorder == 'big':
            data_chunk.byteswap()
        data_all.extend(data_chunk)

        silent = is_silent(data_chunk)

        if audio_started:
            if silent:
                silent_chunks += 1
                if silent_chunks > SILENT_CHUNKS:
                    break
            else: 
                silent_chunks = 0
        elif not silent:
            audio_started = True              

    sample_width = p.get_sample_size(FORMAT)
    stream.stop_stream()
    stream.close()
    p.terminate()

    data_all = trim(data_all)  # we trim before normalize as threshhold applies to un-normalized wave (as well as is_silent() function)
    data_all = normalize(data_all)
    return sample_width, data_all

def record_to_file(path):
    "Records from the microphone and outputs the resulting data to 'path'"
    sample_width, data = record()
    data = pack('<' + ('h' * len(data)), *data)

    wave_file = wave.open(path, 'wb')
    wave_file.setnchannels(CHANNELS)
    wave_file.setsampwidth(sample_width)
    wave_file.setframerate(RATE)
    wave_file.writeframes(data)
    wave_file.close()

if __name__ == '__main__':
    print("Wait in silence to begin recording; wait in silence to terminate")
    record_to_file('demo.wav')
    print("done - result written to demo.wav")

import pyaudio
import wave
from array import array

FORMAT=pyaudio.paInt16
CHANNELS=2
RATE=44100
CHUNK=1024
RECORD_SECONDS=15
FILE_NAME="RECORDING.wav"

audio=pyaudio.PyAudio() #instantiate the pyaudio

#recording prerequisites
stream=audio.open(format=FORMAT,channels=CHANNELS, 
                  rate=RATE,
                  input=True,
                  frames_per_buffer=CHUNK)

#starting recording
frames=[]

for i in range(0,int(RATE/CHUNK*RECORD_SECONDS)):
    data=stream.read(CHUNK)
    data_chunk=array('h',data)
    vol=max(data_chunk)
    if(vol>=500):
        print("something said")
        frames.append(data)
    else:
        print("nothing")
    print("\n")


#end of recording
stream.stop_stream()
stream.close()
audio.terminate()
#writing to file
wavfile=wave.open(FILE_NAME,'wb')
wavfile.setnchannels(CHANNELS)
wavfile.setsampwidth(audio.get_sample_size(FORMAT))
wavfile.setframerate(RATE)
wavfile.writeframes(b''.join(frames))#append frames recorded to file
wavfile.close()

Creo que esto ayudará. Es un script simple que verificará si hay silencio o no, si se detecta silencio no grabará, de lo contrario grabará.

El sitio web de pyaudio tiene muchos ejemplos que son bastante breves y claros: http://people.csail.mit.edu/hubert/pyaudio/

Actualización 14 de diciembre de 2019: ejemplo principal del sitio web vinculado anteriormente de 2017:

"""PyAudio Example: Play a WAVE file."""

import pyaudio
import wave
import sys

CHUNK = 1024

if len(sys.argv) < 2:
    print("Plays a wave file.\n\nUsage: %s filename.wav" % sys.argv[0])
    sys.exit(-1)

wf = wave.open(sys.argv[1], 'rb')

p = pyaudio.PyAudio()

stream = p.open(format=p.get_format_from_width(wf.getsampwidth()),
                channels=wf.getnchannels(),
                rate=wf.getframerate(),
                output=True)

data = wf.readframes(CHUNK)

while data != '':
    stream.write(data)
    data = wf.readframes(CHUNK)

stream.stop_stream()
stream.close()

p.terminate()

Es posible que también desee ver csounds . Tiene varias API, incluido Python. Es posible que pueda interactuar con una interfaz de AD y recopilar muestras de sonido.