如何获得音频文件中特定时间戳的特定频率？

Question

Hi I am currently using Librosa for an audio project I am working on, and I was wondering how can I get the amplitude of a frequency at a specific time-frame in an audio file. I don't know if it is straightforward, but I have looked online and can't find anything. I know that you can produce a spectrogram, but how can you get the information suchas the amplitude of aa frequency at a given timestamp?

EDIT: I meant the amplitude at a timestamp.

Answer 1

The spectrogram is a discrete time-frequency representation. In librosa the frequency bins are along the first axis, and time along the second axis. The frequency bins depend on the number of FFTs chosen, and the time bins depend on the hop length.

Below example shows how to get the amplitude at a given location in the spectrogram, and the associated time and frequency of that location.

import librosa
import numpy

filename = librosa.util.example_audio_file()
y, sr = librosa.load(filename)
n_fft = 1024
hop_length = 512

spec = numpy.abs(librosa.core.stft(y, n_fft=n_fft, hop_length=hop_length))
freqs = librosa.core.fft_frequencies(n_fft=n_fft)
times = librosa.core.frames_to_time(spec[0], sr=sr, n_fft=n_fft, hop_length=hop_length)

print('spectrogram size', spec.shape)

fft_bin = 14
time_idx = 1000

print('freq (Hz)', freqs[fft_bin])
print('time (s)', times[time_idx])
print('amplitude', spec[fft_bin, time_idx])

Similarly you can go from frequency and time to an index in the spectrogram. But since it has been discretized you always have to round to the closest index.