Introduction

The Assistive Listening Device is designed to aid individuals with auditory disabilities by cancelling external noise and boosting frequencies associated with hearing loss. This project utilizes MEMS microphone technology along with Raspberry Pi for audio processing and transmission to an in-ear speaker.

Technologies and Libraries Used

Python

Utilized for implementing signal processing algorithms and data manipulation.

Libraries

• pywavelets and pywt for wavelet transform-based processing.

• numpy for numerical computations.

• scipy for scientific computing and signal processing.

Raspberry Pi 3 Model B+

Serves as the processing unit for audio data. 

IM69D130 MEMS Microphone

Used for capturing audio input.

Methodology

The implementation of the proposed Assistive Listening Device includes the following steps

Audio Collection

Audio is captured using an IM69D130 MEMS microphone. IM69D130 is a high performance digital MEMS microphone making use of Infineon’s Dual Backplate MEMS technology to deliver 105dB dynamic range and high output linearity up to 130dBSPL, with 69 dB(A) signal-to-noise ratio and matched, noise and distortion free audio signals for advanced audio signal processing. 

The collected audio data is then stored in Raspberry Pi 3 Model B+ which acts as the processing unit.

Noise Cancellation

The Assistive Listening Device employs the Wavelet Thresholding Method for noise cancellation, a powerful technique in signal processing. This method involves decomposing the audio signal into different frequency components using wavelet transforms, and then selectively removing noise based on certain thresholding criteria. Wavelet thresholding offers several advantages, including its ability to effectively suppress noise while preserving important signal features. The following types of wavelets are utilized in this method:

Haar Wavelet

It’s a mathematical function used in signal processing and image compression. Haar wavelets are particularly useful for detecting edges and discontinuities in signals, making it suitable for noise cancellation tasks.

this function represents a square wave that takes the value 1 on the interval [0, 1/2) and -1 on the interval [1/2, 1), and it’s zero elsewhere.

Daubechies Wavelets

Daubechies wavelets are a family of orthogonal wavelet transforms. The mathematical function of the Daubechies wavelet, Ψ(x), depends on the specific order of the wavelet (e.g., db1, db2, db3, etc.). For example, the db1 (Haar) wavelet is the same as the Haar wavelet function defined above. Higher-order Daubechies wavelets have more complex mathematical expressions.

Symlets

Symlets, or "Symmetric Daubechies Wavelets”, have improved symmetry properties compared to Daubechies wavelets. The mathematical function of the Symlet wavelet, Ψ(x), depends on the specific order of the wavelet (e.g., sym2, sym3, sym4, etc.).

Bi-orthogonal Wavelets

Bi-orthogonal wavelets use two separate sets of wavelet functions for decomposition and reconstruction. The mathematical functions of the bi-orthogonal wavelet pairs, Ψ(x) and φ(x). These wavelets offer greater flexibility and control over the wavelet transform process.

By leveraging these wavelet functions and the wavelet thresholding method, the Assistive Listening Device effectively removes unwanted noise from the audiosignal, improving the overall auditory experience for users with auditory disabilities.

Frequency Shaping using Filter Banks

In addition to wavelet thresholding, the Assistive Listening Device employs frequency shaping using filter banks. Filter banks are used to divide the audio to signal into different frequency bands, allowing for selective amplification or at tenuation of specific frequency components.

Filter banks consist of multiple bandpass filters arranged in parallel, each covering a specific frequency range. By adjusting the gain of individual filters, the frequency response of the audio signal can be shaped to enhance or suppress certain frequencies.

This frequency shaping technique is particularly useful for boosting frequencies associated with hearing loss while attenuating background noise. By customizing the gain of each filter in the filter bank, the Assistive Listening Device can adapt to the specific auditory needs of the user, providing a personalized listening experience.

Results

Spectrogram of a noisy sound

Spectrogram and Waveform of Denoised Signal

Conclusion

The Assistive Listening Device effectively integrates advanced technologies to address the needs of individuals with auditory disabilities. By leveraging MEMS 3 microphone technology, Python signal processing libraries, and the computational power of the Raspberry Pi, the device enhances the auditory experience by canceling external noise and boosting frequencies. This demonstrates a promising solution for improving accessibility and enhancing the quality of life for users with auditory impairments

References

[1] Bin, Tang, 2005, A study on wireless hearing aids system configuration and simulation, National University of Singapore, oai:scholarbank.nus.edu.sg:10635/15144

[2] Navdeep Kaur, Dr. Hardeep Singh Ryait, 2013, Study Of Digital Hearing Aid Using Frequency Shaping Function, INTERNATIONAL JOURNAL OF ENGINEERING RESEARCH & TECHNOLOGY (IJERT) Volume 02, Issue 05 (May 2013)

[3] M. A. Ali and P. M. Shemi, "An improved method of audio denoising based on wavelet transform," 2015 International Conference on Power, Instrumentation, Control and Computing (PICC), Thrissur, India, 2015, pp. 1-6, doi: 10.1109/PICC.2015.7455802.

[4] I2S MEMS Microphone IM69D130 for Raspberry PI