I have no access to your audio files so I've downloaded: IR from here (mono/r1_omni.wav) - it's a really long one Anechoic recording from here (operatic-voice/mono/singing.wav) Resampled voice ...

From what you've mentioned it looks like the task is for environmental sound event detection. I think that the best starting point for you is to check the DCASE challenge (Detection and Classification ...

I can give you a quick and hacky solution with sox that can be easily installed on any Linux distribution. sox in.wav -n trim 0 0.1 stats : newfile : restart 2>&1 | grep 'RMS lev dB' | awk '{ ...

I did download this annoying video and analyzed the part with a vuvuzela. You should take a note that: This part is clipped The audio is stereo And that pretty much solves the mystery. If you ...

I suggest using Spectral Flatness, aka Wiener Entropy. It is defined as a ratio of geometric and arithmetic mean of the magnitude spectra $X(k)$: $$\Xi=\dfrac{\sqrt[K]{\prod_{k=0}^{K} X(k)}}{\frac{1}{... View answer Accepted answer 3 votes This approach with notch filter with not work. All clicks are impulse-like sounds and we know that an impulse has frequency content at almost every frequency. What you are trying to do, by applying ... View answer 0 votes I am not sure what do you mean by magnitude, but you should be getting a waveform that's perfectly fine for further analysis you wish to do. Just keep in mind that you don't know the frequency ... View answer Accepted answer 19 votes Encouraged by Hilmar, I've decided to update the answer with all the steps necessary to calculate the Reverberation Time from a scratch. Presumably, it will be useful for others interested in this ... View answer Accepted answer 1 votes Let's assume that signal you are analysing is sinusoid with amplitude A: x=a\sin{2\pi f_{0} t} Its RMS value of amplitude is then: \dfrac{a}{\sqrt{2}} as you noticed in your code. Before ... View answer Accepted answer 2 votes Normalizing the filterbanks by their widths is optional and totally up to you (similarly to the warping scale Mel/Bark). Depending on your application, you can start without normalization and see what ... View answer Accepted answer 2 votes First of all, I wouldn't worry too much about the speaker response since it is relatively flat and the microphone has a much bigger roll-off. Since you've captured the frequency response using sweep,... View answer Accepted answer 18 votes Let me start from the beginning. The standard way of calculating cepstrum is following:$$C(x(t))=\mathcal{F}^{-1}[\log(\mathcal{F}[x(t)])] In the case of the MFCC coefficients case is a bit ... View answer 2 votes First of all you should take the magnitude of the FFT (use abs function) - what you've plotted is just a real part of FFT. Secondly, depending on what you want to achieve, I would suggest to detrend ... View answer 4 votes I will answer your questions in reverse order. 4: DTW (Dynamic Time Warping) is not a library but an algorithm. It allows aligning two sequences by warping them in time. You can use it for pretty ... View answer 3 votes Nowadays the easiest thing would be to use librosa for this task. It has the mel_to_stft function which does exactly what you want. As others have mentioned, this reconstruction is lossy and only ... View answer Accepted answer 44 votes One application of the Hilbert Transform is to obtain a so-called Analytic Signal. For signal s(t), its Hilbert Transform \hat{s}(t) is defined as a composition:s_A(t)=s(t)+j\hat{s}(t) $$The ... View answer Accepted answer 10 votes I think it is kind'a similar to soft and hard thresholding using in wavelet de-noising. Have you come across this topic? pywt has already an in-built function for this purpose. Please take a closer ... View answer Accepted answer 18 votes This phenomenon has nothing to do with spectral leakage. What you are observing is the effect of zero padding. Given a number of samples N, there is a maximum possible frequency resolution \Delta f... View answer Accepted answer 11 votes The way MFCC's are always used is by feeding them into the classifier. This can be done on a frame-by-frame basis (12x1 vector), or by concatenating (12xN) - same as a spectrogram. Thus for DTW, you ... View answer Accepted answer 28 votes We know that in general transfer function of a filter is given by:$$H(z)=\dfrac{\sum_{k=0}^{M}b_kz^{-k}}{\sum_{k=0}^{N}a_kz^{-k}}  Now substitute $z=e^{j\omega}$ to evaluate the transfer function ...

This question is very broad and it is not possible to give a very good answer without even knowing the internals of your system (the type of the network and features make a big difference). However, I ...

I will answer your question from my experience. For most of the time, I've been using window length which is a power of 2, trying a few overlap percentages, training the system and picking the one ...

Yes, you can calculate it by summing all of the 3rd-octave band pressures. Keep in mind, that if the units are pressure then simple sum and conversion to the decibel scale is enough. However, if you ...

Personally, I very much like the interactive visualisations of filters that connect various bits together. There is a great website called MicroModeller DSP (I am not affiliated with it). You can ...

If by "its" you mean the value of the fader, then yes - it's absolutely correct. Fader defines the attenuation of the signal with respect to the reference level. The units are in logarithmic scale. ...

You should not be using the analog filter - use a digital filter instead. You want the filter to be defined in Z-domain, not S-domain. Also, you should define the time vector with known sampling ...

Hate to answer this question, but it is a plotting issue, not the actual design problem. Basically, you don't have enough points at low frequencies cause they are spaced linearly and you are plotting ...

It's more like a soft answer (I am happy to update it later), but Alex Acero explained the technology behind aniomoji on this years ICASSP 2018. Here is the link. Basically, they are using so-called ...