import logging import pyaudio import numpy from numpy import pi logging.basicConfig(level=logging.INFO) logger = logging.getLogger(__name__) standard_sample_rates = 1000 * numpy.array([ 8, 9.6, 11.025, 12, 16, 22.05, 24, 32, 44.1, 48, 88.2, 96, 192]) def monitor_pitch(device: int = 5, max_freq: float = 6000, min_freq: float = 10, samples_per_buffer: int = 1024, audio: pyaudio.PyAudio = None, ): if audio is None: audio = pyaudio.PyAudio() supported_sample_rates = [] devinfo = audio.get_device_info_by_index(device) for rate in standard_sample_rates: try: if audio.is_format_supported(rate, input_device=device, input_channels=devinfo['maxInputChannels'], input_format=pyaudio.paInt16): supported_sample_rates.append(rate) except ValueError: pass supported_sample_rates = numpy.array(supported_sample_rates) logger.info('Supported rates: {}'.format(supported_sample_rates)) ''' max_freq < 2 * sample_rate min_freq * 2**(1/12) > freq_resolution (for discrimination), more for accuracy... freq_resolution <= sample_rate / (samples_per_buffer * num_buffers) ''' freq_resolution = min_freq * 2**(1/12) / 10 rate_is_acceptable = supported_sample_rates >= 2 * max_freq sample_rate = int(numpy.min(supported_sample_rates[rate_is_acceptable])) num_buffers = int(numpy.ceil(sample_rate / (samples_per_buffer * freq_resolution))) samples_per_fft = samples_per_buffer * num_buffers logger.info('Running on device {} with {} buffers,'.format(device, num_buffers) + ' {} sample rate, {} samples per buffer'.format( device, num_buffers, sample_rate, samples_per_buffer)) logger.info('Buffers take {:.3g} sec to fully clear'.format(samples_per_fft / sample_rate)) stream = audio.open(format=pyaudio.paInt16, channels=1, rate=sample_rate, input=True, frames_per_buffer=samples_per_buffer) stream.start_stream() # Hanning window window = (1 - numpy.cos(numpy.linspace(0, 2 * pi, samples_per_fft, False))) / 2 freqs = numpy.fft.fftfreq(samples_per_fft, 1 / sample_rate) buf = numpy.zeros(num_buffers * samples_per_buffer, dtype=numpy.float32) note_names = ['C', 'C#', 'D', 'D#', 'E', 'F', 'F#', 'G', 'G#', 'A', 'A#', 'B'] while stream.is_active(): # Shift the buffer down and new data in buf[:-samples_per_buffer] = buf[samples_per_buffer:] buf[-samples_per_buffer:] = numpy.fromstring(stream.read(samples_per_buffer), numpy.int16) fft = numpy.fft.rfft(buf * window) # Get frequency of maximum response in range ind = numpy.abs(fft[1:]).argmax() + 1 freq = freqs[ind] mag = numpy.abs(fft[ind]) # Get note number and nearest note q = numpy.log2(freq/440) n = 12 * q + 69 n0 = int(round(n)) delta = n - n0 logger.info('freq: {:7.2f} Hz mag:{:7.2f} note: {:>3s} {:+.2f}'.format( freq, numpy.log10(mag), note_names[n0 % 12] + str(n0//12 - 1), delta)) delta_part = int(delta // 0.1) if delta_part > 0: signal = ' ' * 6 + '+' * delta_part elif delta_part == 0: signal = ' ' * 5 + '|' elif delta_part < 0: signal = ' ' * (5 + delta_part) + '-' * delta_part logger.info(' {}'.format(signal)) if __name__ == '__main__': audio = pyaudio.PyAudio() logger.info("Available devices:") for device in range(audio.get_device_count()): devinfo = audio.get_device_info_by_index(device) if devinfo['maxInputChannels'] > 0: logger.info('{}: {}'.format(device, devinfo['name'])) monitor_pitch(device=5, min_freq=20)