Global ETD Search

Search theses and dissertations gathered from participating repositories worldwide. Every result links back to the library that holds it. No account is needed.

Results

Showing 1 to 20 of 25 for “"Mel Spectrogram"”.

  1. On Zero-Shot Multi-Speaker Text-to-Speech Using Deep Learning

    … main components: a speaker encoder network, a mel-spectrogram prediction network, and a vocoder network. A speaker encoder network captures the speaker identity in a fixed-sized speaker embedding. This speaker embedding is injected into a mel-spectrogram prediction network at one or more …

    concordia Download PDF for On Zero-Shot Multi-Speaker Text-to-Speech Using Deep Learning (opens in a new tab)

  2. A flexible speech feature converter based on an enhanced architecture of U-net

    … domain signals into various features such as the mel spectral features and WORLD vocoder features. These two types of features can both be extracted from speech or used to synthesize speech. On the other hand, certain applications call for conversion between different types of features. To convert …

    missouri Repository record for A flexible speech feature converter based on an enhanced architecture of U-net (opens in a new tab)

  3. Garsinio signalo atpažinimas /

    … centroid; Short-time Fourier transform based Spectrogram; Mel Spectrogram; Chroma features; Autocorrelation. Short summaries and introductions were also given to three methods of sound classification that can be used for music classification related tasks: Dynamic time warping; Hidden Markov …

    vilnius Repository record for Garsinio signalo atpažinimas / (opens in a new tab)

  4. Linear decoding models of rich audiovisual stimuli utilizing intracranial recordings

    … information of electrodes to the pitch, and the mel-spectrogram of their respective movie trailers. Utilizing an anatomical ablation method, we were able to demonstrate the importance of primary auditory cortical signals in the robustness of our predicted acoustic features. Next, we found …

    texas Repository record for Linear decoding models of rich audiovisual stimuli utilizing intracranial recordings (opens in a new tab)

  5. Intelligent Sound Diagnostics - Acoustic Fault Detection in Truck Steering Systems Using Machine Learning and Deep Learning

    … Seven individual acoustic descriptors, MFCC, log-Mel spectrogram statistics, sub-band energy ratios, spectral shape descriptors, time-domain statistics, amplitude envelope kurtosis, and discrete wavelet transform coefficients, are also evaluated in isolation to determine which is most …

    chalmers Repository record for Intelligent Sound Diagnostics - Acoustic Fault Detection in Truck Steering Systems Using Machine Learning and Deep Learning (opens in a new tab)

  6. A Percussion Method to Classify Bolt Torque for Decommissioned Wind Turbine Blade (DWTB) Structures in the Field

    … algorithms were investigated, including Mel-Frequency Cepstral Coefficients (MFCC) for shallow learning models, K-Nearest Neighbor (KNN), Decision Trees (DT) and Support Vector Machines (SVM) and mel-spectrogram, Short-Time Fourier Transform (STFT), Continuous Wavelet Transform (CWT), …

    houston Repository record for A Percussion Method to Classify Bolt Torque for Decommissioned Wind Turbine Blade (DWTB) Structures in the Field (opens in a new tab) Embargoed until 2027-12-01

  7. Edge Device Speaker Verification

    … system to a microcontroller, we re-produce a log-mel spectrogram framework implemented in Python to our target device supported language: C++. We show that it is possible to build a reasonably accurate,EER ≤11%, generalizable text-independent speaker verification model which will fit on even the …

    cuny Repository record for Edge Device Speaker Verification (opens in a new tab)

  8. Hipoidinės pavaros gedimų šablono atpažinimas naudojant autoenkoderį ir giliuosius neuroninius tinklus /

    … accuracy of 83% using FFT data and 81% using Mel spectrogram data, compared to the 81% accuracy of the CNN-2D model. The autoencoder-based models performed better in classifying fault levels 3 and 4 but worse in classifying fault level 6. The study also explored the performance of these models …

    vilnius Repository record for Hipoidinės pavaros gedimų šablono atpažinimas naudojant autoenkoderį ir giliuosius neuroninius tinklus / (opens in a new tab)

  9. Histogram Matching to Reduce Acoustic Mismatch in Automatic Speech Recognition

    … acoustic mismatch between audio signals. Mel-frequency-dependent histogram matching with a silence threshold used in the log Mel-spectrogram domain is implemented before the decoding step in an automatic speech recognition system. The technique is shown to be effective within a system …

    washington Repository record for Histogram Matching to Reduce Acoustic Mismatch in Automatic Speech Recognition (opens in a new tab)

  10. Respiratory Sound Analysis Using Deep Learning and Machine Learning for Detection of COVID-19 Symptoms

    … the methods of preprocessing the audio into Mel-spectrogram and Mel-frequency cepstrum coefficients (MFCC). Then the paper compared the performance of deep learning and machine learning models with the two different input forms. The deep learning models examined in this research include …

    buffalo Repository record for Respiratory Sound Analysis Using Deep Learning and Machine Learning for Detection of COVID-19 Symptoms (opens in a new tab)

  11. FEW-SHOT VOICE AND FOREIGN ACCENT CONVERSION AND ITS APPLICATIONS IN PRONUNCIATION TRAINING

    … (e.g., from phonetic posteriorgrams) to Mel-spectrogram, conditioned on the target speaker embedding. Moreover, I propose an adversarial training scheme that reduces the speaker-dependent cues from the phonetic posteriorgram. The approach can synthesize more natural speech than …

    tamu Repository record for FEW-SHOT VOICE AND FOREIGN ACCENT CONVERSION AND ITS APPLICATIONS IN PRONUNCIATION TRAINING (opens in a new tab)

  12. Muzikos platformos su žanrų atpažinimo funkcionalumu kūrimas /

    … To achieve this goal, the study employs the mel spectrogram method, which transforms a one-dimensional audio signal into a two-dimensional graphical form that best matches human biological frequency perception. A convolutional neural network (CNN) architecture was implemented for …

    vilnius Repository record for Muzikos platformos su žanrų atpažinimo funkcionalumu kūrimas / (opens in a new tab)

  13. Towards Trustworthy and Robust Multi-modal Machine Learning

    … Parkinson’s detection system using voice and mel-spectrogram data. This system introduces a new bias mitigation technique for CNN-based detection models. In addition, I develop a data-type-agnostic and statistical evaluation framework for model fairness assessments. This framework allows for …

    tamu Repository record for Towards Trustworthy and Robust Multi-modal Machine Learning (opens in a new tab) Embargoed until 2026-12-01

  14. Single and Multi-Label Environmental Sound Classification Using Convolutional Neural Networks

    … patches extracted from the computed mel-spectrogram of the signals. Dropout and weight decay regularization methods are applied and the cross-entropy loss is optimized using Adam algorithm. Results show that these systems are very sensitive to noise and level corruptions of the …

    chalmers Repository record for Single and Multi-Label Environmental Sound Classification Using Convolutional Neural Networks (opens in a new tab)

  15. Diagnosis and prognosis of rolling element bearings using audio signals and deep learning

    … deep convolutional neural network and log-scaled mel-spectrogram-based intelligent system to estimate the stage RUL of bearings. The proposed method is validated using experimental data. and 9 stage classification shows the highest accuracy of 98%.

    edithcowan Repository record for Diagnosis and prognosis of rolling element bearings using audio signals and deep learning (opens in a new tab)

  16. Detecting User Emotions From Audio Conversations With the Smart Assistants

    … an audio. On a concurrent study, I observed that Mel Frequency Cepstral Coefficient (MFCC) in combination with Mel Spectrogram (MEL) result in higher classification accuracy than other existing audio feature combinations available in literature. Upon this conclusion, I have adapted TPOT combined …

    mo-state Repository record for Detecting User Emotions From Audio Conversations With the Smart Assistants (opens in a new tab)

  17. Improving Foreign Accent Conversion with Zero-Shot Learning and Explicit Prosody Modeling

    … input bottleneck features into the corresponding Melspectrograms, conditioned on an embedding from the L2 speaker. During inference, both modules operate in sequence to take an unseen L2 utterance and generate a native-accented Mel-spectrogram. The proposed system achieved a large reduction (67%) …

    tamu Repository record for Improving Foreign Accent Conversion with Zero-Shot Learning and Explicit Prosody Modeling (opens in a new tab)

  18. Specific Loudness Features for Sound Event Classification and Localization in Machine Learning

    … systems that classify and localize sound events. Mel spectrograms are the de factostandard in audio machine learning, yet they are an incomplete perceptual transform: they warp the frequency axis to match human hearing but leave the magnitude axis in raw logpower, omitting the equal-loudness …

    washington Repository record for Specific Loudness Features for Sound Event Classification and Localization in Machine Learning (opens in a new tab)

  19. Design of Multi-modality Deep Fusion Architecture for Deep Acoustic Analytics

    … the multi-features fusion network, such as Log-Mel Spectrogram, Chromagram, and Mel Frequency Cepstral Coefficient. (3) We have developed effective data augmentation and normalization methods to enhance the quality and comparability of sparse audio samples with various extended features. (4) The …

    umkc Repository record for Design of Multi-modality Deep Fusion Architecture for Deep Acoustic Analytics (opens in a new tab)

  20. Audio Based Road Type Classification Using CNNs and AST: Development of Audio Based Road Type Classification Models with Focus on Convolutional Neural Networks and The Audio Spectrogram Transformer Model

    … Networks (CNNs) and the transformer-based Audio Spectrogram Transformer model, in distinguishing between road surface types such as smooth asphalt, rough asphalt and uneven surfaces. Audio data was pre-processed using feature extraction techniques such as Mel-spectrograms and Mel-frequency …

    chalmers Repository record for Audio Based Road Type Classification Using CNNs and AST: Development of Audio Based Road Type Classification Models with Focus on Convolutional Neural Networks and The Audio Spectrogram Transformer Model (opens in a new tab)

Page 1 of 2