Analyzing Noise Robustness of Cochleogram and Mel Spectrogram Features in Deep Learning Based Speaker Recognition
The performance of speaker recognition systems is very well on the datasets without noise and mismatch. However, the performance gets degraded with the environmental noises, channel variation, physical and behavioral changes in speaker. The types of Speaker related feature play crucial role in impro...
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2022-12-01
|
| Series: | Applied Sciences |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2076-3417/13/1/569 |
| _version_ | 1797626165709504512 |
|---|---|
| author | Wondimu Lambamo Ramasamy Srinivasagan Worku Jifara |
| author_facet | Wondimu Lambamo Ramasamy Srinivasagan Worku Jifara |
| author_sort | Wondimu Lambamo |
| collection | DOAJ |
| description | The performance of speaker recognition systems is very well on the datasets without noise and mismatch. However, the performance gets degraded with the environmental noises, channel variation, physical and behavioral changes in speaker. The types of Speaker related feature play crucial role in improving the performance of speaker recognition systems. Gammatone Frequency Cepstral Coefficient (GFCC) features has been widely used to develop robust speaker recognition systems with the conventional machine learning, it achieved better performance compared to Mel Frequency Cepstral Coefficient (MFCC) features in the noisy condition. Recently, deep learning models showed better performance in the speaker recognition compared to conventional machine learning. Most of the previous deep learning-based speaker recognition models has used Mel Spectrogram and similar inputs rather than a handcrafted features like MFCC and GFCC features. However, the performance of the Mel Spectrogram features gets degraded in the high noise ratio and mismatch in the utterances. Similar to Mel Spectrogram, Cochleogram is another important feature for deep learning speaker recognition models. Like GFCC features, Cochleogram represents utterances in Equal Rectangular Band (ERB) scale which is important in noisy condition. However, none of the studies have conducted analysis for noise robustness of Cochleogram and Mel Spectrogram in speaker recognition. In addition, only limited studies have used Cochleogram to develop speech-based models in noisy and mismatch condition using deep learning. In this study, analysis of noise robustness of Cochleogram and Mel Spectrogram features in speaker recognition using deep learning model is conducted at the Signal to Noise Ratio (SNR) level from −5 dB to 20 dB. Experiments are conducted on the VoxCeleb1 and Noise added VoxCeleb1 dataset by using basic 2DCNN, ResNet-50, VGG-16, ECAPA-TDNN and TitaNet Models architectures. The Speaker identification and verification performance of both Cochleogram and Mel Spectrogram is evaluated. The results show that Cochleogram have better performance than Mel Spectrogram in both speaker identification and verification at the noisy and mismatch condition. |
| first_indexed | 2024-03-11T10:06:37Z |
| format | Article |
| id | doaj.art-29fe48e18d3c4f578be1a2a59128cc61 |
| institution | Directory Open Access Journal |
| issn | 2076-3417 |
| language | English |
| last_indexed | 2024-03-11T10:06:37Z |
| publishDate | 2022-12-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Applied Sciences |
| spelling | doaj.art-29fe48e18d3c4f578be1a2a59128cc612023-11-16T14:58:48ZengMDPI AGApplied Sciences2076-34172022-12-0113156910.3390/app13010569Analyzing Noise Robustness of Cochleogram and Mel Spectrogram Features in Deep Learning Based Speaker RecognitionWondimu Lambamo0Ramasamy Srinivasagan1Worku Jifara2Computer Science and Engineering Department, Adama Science and Technology University, Adama P.O. Box 1888, EthiopiaComputer Engineering, King Faisal University, Al Hofuf 31982, Al-Ahsa, Saudi ArabiaComputer Science and Engineering Department, Adama Science and Technology University, Adama P.O. Box 1888, EthiopiaThe performance of speaker recognition systems is very well on the datasets without noise and mismatch. However, the performance gets degraded with the environmental noises, channel variation, physical and behavioral changes in speaker. The types of Speaker related feature play crucial role in improving the performance of speaker recognition systems. Gammatone Frequency Cepstral Coefficient (GFCC) features has been widely used to develop robust speaker recognition systems with the conventional machine learning, it achieved better performance compared to Mel Frequency Cepstral Coefficient (MFCC) features in the noisy condition. Recently, deep learning models showed better performance in the speaker recognition compared to conventional machine learning. Most of the previous deep learning-based speaker recognition models has used Mel Spectrogram and similar inputs rather than a handcrafted features like MFCC and GFCC features. However, the performance of the Mel Spectrogram features gets degraded in the high noise ratio and mismatch in the utterances. Similar to Mel Spectrogram, Cochleogram is another important feature for deep learning speaker recognition models. Like GFCC features, Cochleogram represents utterances in Equal Rectangular Band (ERB) scale which is important in noisy condition. However, none of the studies have conducted analysis for noise robustness of Cochleogram and Mel Spectrogram in speaker recognition. In addition, only limited studies have used Cochleogram to develop speech-based models in noisy and mismatch condition using deep learning. In this study, analysis of noise robustness of Cochleogram and Mel Spectrogram features in speaker recognition using deep learning model is conducted at the Signal to Noise Ratio (SNR) level from −5 dB to 20 dB. Experiments are conducted on the VoxCeleb1 and Noise added VoxCeleb1 dataset by using basic 2DCNN, ResNet-50, VGG-16, ECAPA-TDNN and TitaNet Models architectures. The Speaker identification and verification performance of both Cochleogram and Mel Spectrogram is evaluated. The results show that Cochleogram have better performance than Mel Spectrogram in both speaker identification and verification at the noisy and mismatch condition.https://www.mdpi.com/2076-3417/13/1/569speaker identificationspeaker verificationMel SpectrogramCochleogram2DCNNResNet-50 |
| spellingShingle | Wondimu Lambamo Ramasamy Srinivasagan Worku Jifara Analyzing Noise Robustness of Cochleogram and Mel Spectrogram Features in Deep Learning Based Speaker Recognition Applied Sciences speaker identification speaker verification Mel Spectrogram Cochleogram 2DCNN ResNet-50 |
| title | Analyzing Noise Robustness of Cochleogram and Mel Spectrogram Features in Deep Learning Based Speaker Recognition |
| title_full | Analyzing Noise Robustness of Cochleogram and Mel Spectrogram Features in Deep Learning Based Speaker Recognition |
| title_fullStr | Analyzing Noise Robustness of Cochleogram and Mel Spectrogram Features in Deep Learning Based Speaker Recognition |
| title_full_unstemmed | Analyzing Noise Robustness of Cochleogram and Mel Spectrogram Features in Deep Learning Based Speaker Recognition |
| title_short | Analyzing Noise Robustness of Cochleogram and Mel Spectrogram Features in Deep Learning Based Speaker Recognition |
| title_sort | analyzing noise robustness of cochleogram and mel spectrogram features in deep learning based speaker recognition |
| topic | speaker identification speaker verification Mel Spectrogram Cochleogram 2DCNN ResNet-50 |
| url | https://www.mdpi.com/2076-3417/13/1/569 |
| work_keys_str_mv | AT wondimulambamo analyzingnoiserobustnessofcochleogramandmelspectrogramfeaturesindeeplearningbasedspeakerrecognition AT ramasamysrinivasagan analyzingnoiserobustnessofcochleogramandmelspectrogramfeaturesindeeplearningbasedspeakerrecognition AT workujifara analyzingnoiserobustnessofcochleogramandmelspectrogramfeaturesindeeplearningbasedspeakerrecognition |