我们从Python开源项目中，提取了以下7个代码示例，用于说明如何使用librosa.magphase()。
def griffin_lim(mag, phase_angle, n_fft, hop, num_iters): """Iterative algorithm for phase retrival from a magnitude spectrogram. Args: mag: Magnitude spectrogram. phase_angle: Initial condition for phase. n_fft: Size of the FFT. hop: Stride of FFT. Defaults to n_fft/2. num_iters: Griffin-Lim iterations to perform. Returns: audio: 1-D array of float32 sound samples. """ fft_config = dict(n_fft=n_fft, win_length=n_fft, hop_length=hop, center=True) ifft_config = dict(win_length=n_fft, hop_length=hop, center=True) complex_specgram = inv_magphase(mag, phase_angle) for i in range(num_iters): audio = librosa.istft(complex_specgram, **ifft_config) if i != num_iters - 1: complex_specgram = librosa.stft(audio, **fft_config) _, phase = librosa.magphase(complex_specgram) phase_angle = np.angle(phase) complex_specgram = inv_magphase(mag, phase_angle) return audio
def transform_audio(self, y): '''Compute the STFT magnitude and phase. Parameters ---------- y : np.ndarray The audio buffer Returns ------- data : dict data['mag'] : np.ndarray, shape=(n_frames, 1 + n_fft//2) STFT magnitude data['phase'] : np.ndarray, shape=(n_frames, 1 + n_fft//2) STFT phase ''' n_frames = self.n_frames(get_duration(y=y, sr=self.sr)) D = stft(y, hop_length=self.hop_length, n_fft=self.n_fft) D = fix_length(D, n_frames) mag, phase = magphase(D) if self.log: mag = amplitude_to_db(mag, ref=np.max) return {'mag': mag.T[self.idx].astype(np.float32), 'phase': np.angle(phase.T)[self.idx].astype(np.float32)}
def transform_audio(self, y): '''Compute the CQT Parameters ---------- y : np.ndarray The audio buffer Returns ------- data : dict data['mag'] : np.ndarray, shape = (n_frames, n_bins) The CQT magnitude data['phase']: np.ndarray, shape = mag.shape The CQT phase ''' n_frames = self.n_frames(get_duration(y=y, sr=self.sr)) C = cqt(y=y, sr=self.sr, hop_length=self.hop_length, fmin=self.fmin, n_bins=(self.n_octaves * self.over_sample * 12), bins_per_octave=(self.over_sample * 12)) C = fix_length(C, n_frames) cqtm, phase = magphase(C) if self.log: cqtm = amplitude_to_db(cqtm, ref=np.max) return {'mag': cqtm.T.astype(np.float32)[self.idx], 'phase': np.angle(phase).T.astype(np.float32)[self.idx]}
def transform_audio(self, y): '''Compute the HCQT Parameters ---------- y : np.ndarray The audio buffer Returns ------- data : dict data['mag'] : np.ndarray, shape = (n_frames, n_bins, n_harmonics) The CQT magnitude data['phase']: np.ndarray, shape = mag.shape The CQT phase ''' cqtm, phase = [], [] n_frames = self.n_frames(get_duration(y=y, sr=self.sr)) for h in self.harmonics: C = cqt(y=y, sr=self.sr, hop_length=self.hop_length, fmin=self.fmin * h, n_bins=(self.n_octaves * self.over_sample * 12), bins_per_octave=(self.over_sample * 12)) C = fix_length(C, n_frames) C, P = magphase(C) if self.log: C = amplitude_to_db(C, ref=np.max) cqtm.append(C) phase.append(P) cqtm = np.asarray(cqtm).astype(np.float32) phase = np.angle(np.asarray(phase)).astype(np.float32) return {'mag': self._index(cqtm), 'phase': self._index(phase)}
def parse_audio(self, audio_path): if self.augment: y = load_randomly_augmented_audio(audio_path, self.sample_rate) else: y = load_audio(audio_path) if self.noiseInjector: add_noise = np.random.binomial(1, self.noise_prob) if add_noise: y = self.noiseInjector.inject_noise(y) n_fft = int(self.sample_rate * self.window_size) win_length = n_fft hop_length = int(self.sample_rate * self.window_stride) # STFT D = librosa.stft(y, n_fft=n_fft, hop_length=hop_length, win_length=win_length, window=self.window) spect, phase = librosa.magphase(D) # S = log(S+1) spect = np.log1p(spect) spect = torch.FloatTensor(spect) if self.normalize: mean = spect.mean() std = spect.std() spect.add_(-mean) spect.div_(std) return spect
def parse_audio(self, audio_path): if self.augment: y = load_randomly_augmented_audio(audio_path) else: y = load_audio(audio_path) if self.noiseInjector: add_noise = np.random.binomial(1, self.noise_prob) if add_noise: y = self.noiseInjector.inject_noise(y) n_fft = int(self.sample_rate * self.window_size) win_length = n_fft hop_length = int(self.sample_rate * self.window_stride) # STFT D = librosa.stft(y, n_fft=n_fft, hop_length=hop_length, win_length=win_length, window=self.window) spect, phase = librosa.magphase(D) # S = log(S+1) spect = np.log1p(spect) spect = torch.FloatTensor(spect) if self.normalize: mean = spect.mean() std = spect.std() spect.add_(-mean) spect.div_(std) return spect
