Inferred Lattice Recursive Least Squares Filter with DCT-Hadamard Pre-Processing Technique for Speech Enhancement in Digital Hearing Aid

Digital hearing aids is an important support for hearing impaired person. In real-world scenarios, hearing aids often encounter various environmental noise sources that can interfere with speech perception. The proposed work addresses this issue using the Inferred Lattice Recursive Least Squares (ILRLS) algorithm for speech enhancement in hearing aid applications. ILRLS algorithm is applied to the context of digital hearing aids with the objective of enhancing the speech signal while suppressing background noise. Speech samples are collected for the NOIZEUS, SpEAR and TIMIT database for experimentation and validation. The speech samples are pre-processed using Discrete Cosine Transform - Hadamard hybrid algorithm to arrange samples with respect to its frequency components, and to adapt in the algorithm effectively. Effectiveness of the proposed work is assessed by metrics like signal-to-noise ratio, perceptual evaluation of speech Quality, and Mean Square Error. Subjective listening tests were also conducted to gather user feedback. The proposed method significantly enhances both speech quality and intelligibility in noisy environments. The adaptive nature of ILRLS allows it to adapt to changing noise conditions, making it well-suited for digital hearing aids