Combined Bidirectional Long Short-Term Memory with Mel-Frequency Cepstral Coefficients Using Autoencoder for Speaker Recognition

Recently, neural network technology has shown remarkable progress in speech recognition, including word classification, emotion and identity recognition. This paper introduces three novel speaker recognition methods to improve accuracy. The first method, called long short-term memory with mel-frequency cepstral coefficients for triplet loss (LSTM-MFCC-TL), utilizes MFCC as input features the LSTM model incorporates cluster training effective training. second bidirectional (BLSTM-MFCC-TL), enhances accuracy by employing a model. third autoencoder (BLSTM-MFCCAE-TL), an extract additional AE features, which are then concatenated fed into BLSTM results showed that performance of was superior model, method adding achieved best learning effect. Moreover, proposed exhibit faster computation times compared reference GMM-HMM Therefore, utilizing pre-trained autoencoders encoding obtaining can significantly enhance Additionally, it also offers time traditional methods.