# Tag Info

28

Does the Nyquist frequency of the Cochlear nerve impose the fundamental limit on human hearing? No. A quick run-through the human auditory system: The outer ear (pinnae, ear canal), spatially "encodes" the sound direction of incidence and funnel the sound pressure towards the ear drum, which converts sound into physical motions, i.e. mechanical ...

17

It is really quite simple for values between 1 and -1: valueDBFS = 20*log10(abs(value)) If you have values between another range, for example 16bit, then it's: valueDBFS = 20*log10(abs(value)/32768) (because 16bit signed has values between -32768 and +32767, e.g. 2^15=32768) And so the general formula is: valueDBFS = 20*log10(abs(value)/maxValue) The ...

17

The closest example I can think of is the beginning of Suzanne Vega's "Tom's Diner" which has been used for the mpeg-1 layer 3 development, and is still occasionally used to demo audio codecs.

16

Audio processing is a large field, but specifically in speech processing, an open database of samples known as Harvard Sentences is widely used. Harvard sentences are phonetically balanced collections of sentences in American English. Many equivalent databases exist for other languages as well. Actual files with Harvard Sentences speech can be found here.

15

No. Why do you think it would? First of all, the human brain works very different then any human constructed computer (to date); so the assumption that it runs mathematical "algorithms" is somewhat dicey.. Here is roughly how it works: Sound wiggles the air drum That vibration is transferred by the ossicles to the cochlea. The ossicles act mainly as a ...

12

The FFT can only be performed over a limited chunk of data. The basic math is based on the assumption that the time domain signal is periodic, i.e. your chunk of data is repeated in time. That typically results in a major discontinuity at the edges of the chunk. Let's look at a quick example: FFT size = 1000 points, Sample Rate = 1000 Hz, Frequency ...

11

What you want to do is called "Super Resolution" in the field of imaging. This is an ill posed problem. Hence in order to solve it you need some prior / model about your audio data. For example you may look at the model in Speech Super Resolution Using Parallel WaveNet.

9

You're running into a property of the DFT that is usually used in the opposite direction: stuffing zeros between samples in one domain results in replication of the entire sequence in the opposite domain. Let's start in the frequency domain with the signal you plotted, $X[k]$, and look at the inverse transform to see what time-domain signal it corresponds to....

9

While I don't know about tools for upsampling using AI, I am challenging the assumption that the main problem with your sample is lost high frequncies due to resampling from 44kHz to 22kHz, so "just" upsampling is unlikely to solve your root problem. The possibility of AI upsampling itself is sensible, though: While standard (mathematical) ...

8

As you have realized, the hard part of doing digital communications is carrier, symbol and frame synchronization, and channel estimation/equalization. The bad news is that you can't get around these problems. The good news is that implementing these is not that hard, as long as you limit yourself to narrowband BPSK. I know, because I have done this myself, ...

7

The beat and onset detection algorithms used at the Echo Nest are probably variants/improvements of the techniques developed by Tristan Jehan in his Ph.D. This is not the only approach, and I would recommend you to try first: Getting an onset detection function using spectral flux or complex amplitude. Using this algorithm to detect beats (you can improve ...

7

I believe that this "color graph" you are looking for is a spectrogram (although it looks to me more like a scalogram, but you did not mentioned wavelets). Let me give you an example in MATLAB of obtaining such plot: load handel nfft = 512; noverlap = 128; win = hamming(nfft); spectrogram(y, win, noverlap, nfft, Fs, 'yaxis') colormap('jet') So first line ...

7

From the ones I've been using I can recommend: YAAFE - very pleasant to work with in Python ESSENTIA - another one I like particularly due to Python integration aubio FEAPI Aquila - friend of mine used it extensively and he likes it a lot Recently I came across this paper and I believe that this should perfectly answer your question. Moffat D. et al - ...

7

In line with a previous similar question here are my suggestions: There are so many nice books but I believe you should first have a look at the science of sound from Rossing for getting the most broad view on the subject. Then you can look at the following books , each dealing with a particular dimension of the problem: 1- Acoustics_BERANEK 2- Elements ...

7

An exponentially decaying envelope $a\exp(-b x)$ is a good choice, and is used for example in vintage Yamaha FM synthesizers. It has the favorable property that over any constant length time interval, by the end of the interval the envelope has decayed to a constant fraction of what it was at the beginning of the interval. Damped oscillation (with some ...

6

Nonlinear systems are very hard to classify and there's no unified theory like for linear systems. In general, you cannot measure/identify non-linear systems in finite time. There are some specific classes of nonlinear systems which allow for identification or at least approximation. You already named a trivial one: The memoryless nonlinear system. In ...

6

The matlab codes that implemented D. Ellis's algorithm are on their website: http://labrosa.ee.columbia.edu/projects/beattrack/

6

"Ok Google" is described in many publications by Google Automatic Gain Control and Multi-style Training for Robust Small-Footprint Keyword Spotting with Deep Neural Networks Convolutional Neural Networks for Small-Footprint Keyword Spotting It is based on DNN specifically trained for keyphrase and runs really fast. It does not consume a lot of power even ...

6

The main reason why the tables don't have it, is that it's hard to measure. The measurement technique in ISO 354:2003 relies on measuring the difference in reverberation times in a reverberation rooms with and with/out a material sample. At higher frequencies, the reverb time is dominated by air absorption and and the sound field becomes less and less ...

6

My own interpretation: Sound: a mechanical wave that propagates through the air or water. Audio: sound in the 20 Hz to 20 kHz range; in other words, sound that is (at least in theory) audible to a large number of humans. Voice: sound produced by the human vocal tract. Speech: intelligible voice (i.e. not grunts or screams) Tone: signal dominated by a ...

5

I have tried the following: Launch Audacity. Generate a 15000 Hz tone in the track created by default. Add a new track. Generate a 15400 Hz tone in the new track. A lower frequency tone appears during playback. The reason is that both tracks have high levels, so their sum exceeds 1.0; and Audacity applies clipping or limiting. This non-linear operation is ...

5

I also think that NIST database is very popular when it comes to speech recognition tasks. In fact it is a standard for comparison of new algorithms and techniques during yearly challenges. Additionally MATLAB's load handel containing snippet of G.F.Handel - Hallelujah is quite commonly used.

5

Eventhough the exact cause of those wide-spread line spectra is not very clear to me from the supplied information, it's most probably due to the on-off switching implied by the silence period you have added in between your message signals. The on-off waveform is an implicit operation you multiply to your signal. Which has a value of $1$ during the sound ...

5

i don't consider this a "bad" question. But there is a lot that nano needs to deal with. first, you must be able to think about conceptually and mathematically converting your continuous-time signal $$x(t) = \sin(\Omega t)$$ into a discrete-time signal $$x[n] = \sin(\omega n)$$ how $n$ is related to $t$ and how $\omega$ is related to $\Omega$. ...

5

As your plot shows, the second form allows for the correlation peak to be negative. Now, what does a strong negative cross correlation mean? It means the signals are very similar, except one has a negative sign in front of it, i.e., $x_1 \approx -x_2$. Whether or not this makes sense depends a lot on the actual application. In the application you describe, ...

4

There is a huge body of research and commercial work on doing acoustic simulation. There are a two basic classes of simulation systems: Finite element or finite boundary: these model the actual physics by dividing air and or/surfaces into small patches and modelling step by step the interaction between the patches by locally solving the wave equations ...

4

All the standards define dBFS as an RMS measurement, relative to the RMS level of a full-scale sine wave, so the calculation is: value_dBFS = 20*log10(rms(signal) * sqrt(2)) = 20*log10(rms(signal)) + 3.0103 A full-scale sine wave is 0 dBFS A full-scale square wave is +3 dBFS The similar unit dBov is defined in relation to power ratios (so it's also an RMS ...

4

The European Broadcasting Union's (EBU) Sound Quality Assessment Material (SQAM) resource is pretty popular. https://tech.ebu.ch/publications/sqamcd

4

What you are observing is the digital representation of the voltage, which in fact represents the acoustic pressure. Workflow would be something like: Vibrating larynx is producing Acoustic Pressure [Pa] Variations of that pressure are converted by the microphone into Voltage [V] Voltage is being sampled and quantized by the ADC (Analog to Digital Converter)...

4

When using a constant tone audio beacon, beware of room echoes causing multi-path interference and distortion, especially around the leading and trailing portions of your received waveforms. Try using a frequency sweep instead of a constant tone for your transmit waveform. This might provide you with a sharper correlation peak that is less likely to have ...

Only top voted, non community-wiki answers of a minimum length are eligible