Skip to content
Free planNo credit card requiredStart writing
AI for Research home
AI for Research home
Dashboard
Sign inGet started

Search 250 million research papers

Read the abstract, see who wrote it and who cites it, and get the citation in APA, MLA, Chicago or Harvard.

About 10,000 papers for “Spectrogram”

Amplitude and Spectrogram of advertisement calls.

K. S. Seshadri (2558125), Ramit Singal (2558113), H. Priti (2550478) et al. · Figshare · 2016

<p>(a) Advertisement call of <i>M</i>. <i>laterite</i> sp. nov. A single call of 0.754s duration having 103 pulses, (b) a dominant frequency of 3664 Hz and 2<sup>nd</sup> harmonics of 6506 Hz. (c) Advertisement call of <i>M</i>. <i>sholigari</i>. A single call of 0.726 s duration having 70 pulses and (d) a dominant frequency of 3518 Hz.</p>

Cited by 0Open access

LINEAR TAG PREDICTION FROM SPECTROGRAM COVARIANCE

James Bergstra, Michael M, Razvan Pascanu et al. · 2013

features as an alternative to means and variances in MFCCs for condensing spectral features for linear classification. These features are quick and easy to compute (pseudo-code given below) work well with linear classification. Together they offer a viable approach to web-scale music classification, and a competitive null model for research on new datasets. 1. FEATURES We follow [1] in distinguishing two levels, or stages, of feature extraction. Firstly, at the frame level (typically 20-50 milliseconds) we extract features such as Mel Frequencyscale Cepstreal Coefficients (MFCCs). Secondly, at the segment level (roughly 4 seconds) we summarize framelevel features via mean and covariance statistics. For our frame-level features we partitioned the audio into contiguous

Cited by 0

MAE-AST: Masked Autoencoding Audio Spectrogram Transformer

Alan Baade, Puyuan Peng, David Harwath · Interspeech 2022 · 2022

In this paper, we propose a simple yet powerful improvement over the recent Self-Supervised Audio Spectrogram Transformer (SSAST) model for speech and audio classification.Specifically, we leverage the insight that the SSAST uses a very high masking ratio (75%) during pretraining, meaning that the vast majority of self-attention compute is performed on mask tokens.We address this by integrating the encoder-decoder architecture from Masked Autoencoders are Scalable Vision Learners (MAE) into the SSAST, where a deep encoder operates on only unmasked input, and a shallow decoder operates on encoder outputs and mask tokens.We find that MAE-like pretraining can provide a 3× speedup and 2× memory usage reduction over the vanilla SSAST using current audio pretraining strategies with ordinary model and input sizes.When finetuning on downstream tasks, which only uses the encoder, we find that our approach outperforms the SSAST on a variety of downstream tasks.We further conduct comprehensive evaluations into different strategies of pretraining and explore differences in MAE-style pretraining between the visual and audio domains.

Cited by 76

Correcting Mispronunciations in Speech using Spectrogram Inpainting

Talia Ben Simon, Felix Kreuk, Faten Awwad et al. · Interspeech 2022 · 2022

Learning a new language involves constantly comparing speech productions with reference productions from the environment.Early in speech acquisition, children make articulatory adjustments to match their caregivers' speech.Grownup learners of a language tweak their speech to match the tutor reference.This paper proposes a method to synthetically generate correct pronunciation feedback given incorrect production.Furthermore, our aim is to generate the corrected production while maintaining the speaker's original voice.The system prompts the user to pronounce a phrase.The speech is recorded, and the samples associated with the inaccurate phoneme are masked with zeros.This waveform serves as an input to a speech generator, implemented as a deep learning inpainting system with a U-net architecture, and trained to output a reconstructed speech.The training set is composed of unimpaired proper speech examples, and the generator is trained to reconstruct the original proper speech.We evaluated the performance of our system on phoneme replacement of minimal pair words of English as well as on children with pronunciation disorders.Results suggest that human listeners slightly prefer our generated speech over a smoothed replacement of the inaccurate phoneme with a production of a different speaker.

Cited by 1

Spectrogram-based formant tracking via particle filters

Yu Shi, Eric Chang · 2003 IEEE International Conference on Acoustics, Speech, and Signal Processing, 2003. Proceedings. (ICASSP '03). · 2003

The paper presents a particle-filtering method for estimating formant frequencies of speech signals from spectrograms. First, frequency bands corresponding to the analyzed formants are extracted via a two-step dynamic programming based algorithm. A particle-filtering method is then used to locate accurately formants in every formant area based on the posterior PDF described by a set of support points with associated weights. Formant trajectories of voiced frames of a group of 81 utterances were manually tracked and labeled, partly for model training and partly for algorithm evaluation. In the experiments, the proposed method obtains average estimation errors of 72, 115, and 113 Hz for the first three formants, respectively, whereas the LPC based method induces 118, 172, and 250 Hz deviations. The experimental results show that the formants estimated by the proposed method are quite reliable and the trajectories are more accurate than LPC.

Cited by 25

Spectrogram representation of the six whistle types.

Weijie Fu (11877974), Zhongchang Song (719628), Wenjie Xiang (21094912) et al. · Figshare · 2025

<b>(A)</b> constant, <b>(B)</b> upsweep, <b>(C)</b> downsweep, <b>(D)</b> concave, <b>(E)</b> convex, and <b>(F)</b> sinusoidal.

Cited by 0Open access

DE-A PWI SFR/LFC Spectrogram Plots.

Donald A. Gurnett · Open MIND · 2026

This dataset contains dynamic spectrogram PNG plots of the DE-1/PWI Step Frequency Correlator and Low Frequency Correlator data. Each plot spans one orbital period (6 hours 49 minutes). There are three types of files in this dataset. Those derived from data using: 1) the spin-plane 200m electric antenna (Ex), 2) the spin-axis 8m electric antenna (Ez) and 3) the magnetic antennas (B-H) consisting of a 1.0m^2 loop antenna in the spin plane and a search coil on the spin axis. Each image consists of two panels. The title above each panel indicates the instrument, antenna and frequency range. Each panel is a plot of the power spectral density of received signal (color scale) as a function of operating frequency (in a logarithmic scale on the vertical axis) and time (horizontal axis). Beneath the time labels on the horizontal axis of the spectrograms are ephemeris data: position of the spacecraft in radial distance (Earth radii), McIlwain L-shell, magnetic local time, and geomagnetic latitude. Overlaid on each image are traces of the electron cyclotron frequencies. The file naming convention is: de1_pwi_0000_YYYYMMDD_HHMM_AAA.png where: 0000 - Replace with orbit number YYYYMMDD - Replace with date HHMM - Replace with start time AAA - Replace with antenna string

Cited by 0Open access

SepTr: Separable Transformer for Audio Spectrogram Processing

Nicolae-Cătălin Ristea, Radu Tudor Ionescu, Fahad Shahbaz Khan · Interspeech 2022 · 2022

Following the successful application of vision transformers in multiple computer vision tasks, these models have drawn the attention of the signal processing community. This is because signals are often represented as spectrograms (e.g. through Discrete Fourier Transform) which can be directly provided as input to vision transformers. However, naively applying transformers to spectrograms is suboptimal. Since the axes represent distinct dimensions, i.e. frequency and time, we argue that a better approach is to separate the attention dedicated to each axis. To this end, we propose the Separable Transformer (SepTr), an architecture that employs two transformer blocks in a sequential manner, the first attending to tokens within the same time interval, and the second attending to tokens within the same frequency bin. We conduct experiments on three benchmark data sets, showing that our separable architecture outperforms conventional vision transformers and other state-of-the-art methods. Unlike standard transformers, SepTr linearly scales the number of trainable parameters with the input size, thus having a lower memory footprint. Our code is available as open source at https://github.com/ristea/septr.

Cited by 2Open access

Optimality property of the Gaussian window spectrogram

A. J. E. M. Janssen · IEEE Transactions on Signal Processing · 1991

It is shown that for any signal x(t) the minimum of integral /sub - infinity //sup infinity / integral /sub - infinity //sup infinity / ((t-t/sub x/)/sup 2/+(f-f/sub x/)/sup 2/) S/sub x//sup (w)/(t, f) dt df over all normalized time-windows w(t) is achieved by the Gaussian window w(t)=2/sup 1/4/ exp (- pi t/sup 2/). Here (t/sub x/, f/sub x/) is the center of gravity of the signal x(t), S/sub x//sup (w)/ (t, f) is the spectrogram of x(t) due to the window w(t), and the double integral is a measure of the spread of S/sub x//sup (w)/ (t, f) around (t/sub x/, f/sub X/) in the time-frequency plane.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>

Cited by 56Open access

Reassigned spectrogram and its time-domain conversion

Tianyi He · IDEALS (University of Illinois Urbana-Champaign) · 2016

Audio information accounts for a large portion of existing digital data, and numerous researchers are constantly developing new methods and tools for audio analysis. Short-time Fourier transformed based analysis is one of the most prominent analysis tools because of its linearity; however, this mechanism cannot resolve issues with inaccurate localization of energy in both time and frequency, especially for audio editing types of applications. Thus, the reassigned spectrogram method, developed by Auger and Flandrin in 1995, has won its place due to higher accuracy in energy localization by turning discretely sampled signal into continuous domain, and by reassigning the energy to the center of mass for each analysis bin. This method, nevertheless, is non-linear, and it is very di cult to synthesize analysis data back to time-domain representation. This thesis introduces mechanisms to solve this problem and summarizes the result of testing.

Cited by 0Open access

Real-Time Colored Display of Sound Spectrogram

Yasuaki Nakano, Takeshi Nakayama, Tsuneji Koshikawa et al. · The Journal of the Acoustical Society of America · 1967

If signal spectrograms are displayed in real time, so-called brightness modulation display is used in general. Because of the limited dynamic range of brightness modulation of CRT, details of the spectrum are lost very often. In order to overcome such defects and be able to observe instantaneous spectrum patterns more clearly and easily, we constructed new display apparatus in which effective use of visual function is considered. In this system, instantaneous spectrum amplitude is quantized in 11 levels with the step of 4 dB, and each amplitude level is converted to predetermined hue resembled to contour map. Furthermore, these colored instantaneous spectrograms are memorized in sky-sign like fashion in a certain time interval. As the results of displaying various acoustic signals, it is confirmed that color display of this type is more useful than monochromatic display for the visual discrimination of the characteristics corresponding to each signal differences.

Cited by 0Open access

Spectrogram Analysis for Low-Cost Subsurface Detection

Monica Magallon-Mariscal, Danilo Cáceres-Hernández, Ricardo Gutiérrez · 2025

The detection of objects buried in the subsurface, from radar signals, has been studied in many contexts, such as military applications, geological and environmental uses, and many more. In this work, we use the CDM324 Doppler radar, a low-cost sensor that allows us to capture variations in the signal reflected by objects under the surface. The methodology is based on signal processing to generate spectrograms, which allow us to visualize the spectral response of varied materials. Then, by comparing spectral coincidences between multiple records, we seek to identify characteristic patterns that differentiate elements such as aluminum, iron, and brass, as well as non-metallic materials, including soil and organic remains.

Cited by 0

Representation of complex spectrogram via phase conversion

Kohei Yatabe, Yoshiki Masuyama, Tsubasa Kusano et al. · Nippon Onkyo Gakkaishi/Acoustical science and technology/Nihon Onkyo Gakkaishi · 2019

As importance of the phase of complex spectrogram has been recognized widely, many techniques have been proposed for handling it. However, several definitions and terminologies for the same concept can be found in the literature, which has confused beginners. In this paper, two major definitions of the short-time Fourier transform and their phase conventions are summarized to alleviate such complication. A phase-aware signal-processing scheme based on phase conversion is also introduced with a set of executable MATLAB functions (https://doi.org/10/c3qb).

Cited by 34Open access