Projects The life of a busy man... My projects range from those required by school, those done in conjunction with school but not required, those freelance job related, and those that are quite simply personal. Search M-Lester.com Search Google Type of Project Status Description Internship, Shure Inc. Completed I can't tell you yet!...But I will once the product is released! Undergraduate Project Completed Using Multi-rate Filter Banks for a Multi-Band Dynamic Range Compression Plug-in A dynamic compressor attenuates the strong sections of a signal in order to reduce the dynamic range of the signal. Most of the compressors used for music or speech production operate on the full-band principle. The entire audio signal is processed via a single gain-control element as described above. A multi-band compressor includes a set of filters that splits the audio signal into two or more frequency bands. Three, four, or five band compressors are perhaps the best compromise between versatility and computational simplicity. After passing through the filters, each frequency band is fed into its own compressing algorithm. After processing, the signals are recombined. This project employs a tree structured multi-rate filter bank to separate the frequency response of a system into 4 frequency bands so that each band can be compressed separately. The Project Write-up can be found here - Word Document (2MB) The Project Presentation can be found here - Powerpoint Document (2MB) Undergraduate Project Completed Implementation of the Karplus-Strong String Simulation Algorithm on an Analog Devices ADSP processor This involves using a comb filter that is lowpass reverberated to produce the harmonic structure of plucked string. An all pass filter is used to alter the delay of the comb filter with a resolution that can be a fraction sample delay in order to tune the simulation to any pitch. Undergraduate Independent Study Suspended ELJA - Electric-stringed Jazz Pedagogy Jazz improvisation can be a very complicated or a very simple concept depending on your method to learning and interacting. Some people are auditory learners; some visual or otherwise. This orientation directly affects the method of learning jazz improvisation. This project allows the user to bridge the gap between improvisation conceptualization and the physical stringed instrument. By allowing the instrument to physically show the improvisational concepts, the difficulty in connecting the mental to the physical interface is reduced. This project allows the immediate utility of beginner and intermediate improvisation concepts such as chord/scale relation and common tones; however, future expansion also lends itself to the generation and analysis of more advanced concepts such as motifs and "what you didn't play instead of what you did play." An electric bass outfitted with LEDs in the neck are controlled by your computer. The user enters into a computer program the tempo, form, and chords of a jazz tune (or selects a predefined one). The program then calculates the appropriate chord/scales for the tune and allows the user to make any desired corrections. When the user agrees on what I call the "improvisation map" the program counts off the tempo and starts the song. One color LED shows the scale for each chord, while another is user-programmable to functions such as highlighting certain notes such as 3rds and 7ths or to show common tones with the next chord. The music that the user plays is sent back to the computer to be recorded for playback and analyzed . Using typical jazz improvisation techniques, the computer analyzes the music and display information to the user such as "right" and "wrong" notes, notes that the user plays too much or little, approximate calculation of the user's 'style' (i.e. bebop blues vs. straightforward 12 bar blues) , or suggestions of artists to listen to based on the user's 'style.' The possible uses of the feedback element to the computer are unbounded. Please see the following project write-up: ELJA Writeup Graduate Project In Progress JPEG CODEC implementation on a Motorola 56002 DSP When the project is complete I will comment on it. Undergraduate Project Completed Isolated Speech Recognition This project for my Speech and Audio Processing class uses Dynamic Time Warping to do Speech Recognition on the numbers 0-9 for a static speaker. Check out the page here. Undergraduate Project Completed DSP Audio Algorithm implementation on Analog Devices DSP This project includes implementations of typical Audio DSP applications such as volume, panning, and FIR and IIR filters. Undergraduate Project Completed Investigation of the Phonemes in the English Language This project for my Speech and Audio Processing class investigates the spectra and temporal characteristics of the Phonemes in the English language. The analysis is done in Matlab. Check out the page here. Undergraduate Project Completed Speech Production of the Male and Female speaking Vowels This project for my Speech and Audio Processing class uses MATLAB to generate the vowels used in speech. Check out the page here. Proposed Graduate School Project Proposed MaxiReal - Maximum-gain knob Reality emulator This project will most likely be a VST plug-in to be used by studio engineers to aid them in accurate recording of what the ear actually hears. It takes into consideration the Fletcher Munson curves of hearing perception which I believe is a common cause for inaccurate-sounding recordings of loud instruments such as distorted guitar. Many musicians only like the way their guitar sounds at volume 10. I hypothesize that part of the reason, beyond the physical and electronic overdriving which can in fact be accurately recorded, is because at these high SPL levels our ear reacts differently than if the sound source was soft. A sound recorded at 110+ dB SPL then monitored at 60 dB or less, for instance, will not sound accurate because between the two SPL levels there is no 'ear' mechanism which colors the sound. By applying the Fletcher Munson equalization curves on a DSP, I hope to provide studio engineers with a more accurate recording option. Proposed Graduate School Project Proposed Audio Compression using Asymmetric Sampling The Nyquist Theorem states that in order to accurately capture and reproduce a signal you must sample it at at least twice the highest frequency component present in the signal. In audio, we assume that the highest frequency a human can hear is 20KHz, so we sample at or above that rate, typically at 44.1 KHZ. However, if there are no high frequencies present in the system -- moreover if there are no important (audible) high frequencies, then sampling at that rate is not necessary. The unnecessary information that is coded is redundant, one of the very things compression is trying to eliminate. If the sampling rate was able to adaptively change depending on the highest frequency present in the signal, higher compression rates could be attained. An example besides typical storage concerns of audio such as wav, mp3, etc. is that of satellite or other streaming audio. Throughout a broadcast different songs or other program content are streamed; each type of song or content has a different required sampling rate. For example, if there is a DJ announcing something between songs, it is unnecessary to code it at 44.1 KHz. Undergraduate Project Complete Adaptation and Improvement of the MPEG Model 1 Standa
Projects
The life of a busy man...
My projects range from those required by school, those done in conjunction with school but not required, those freelance job related, and those that are quite simply personal.
Type of Project
Status
Description
...But I will once the product is released!
Using Multi-rate Filter Banks for a Multi-Band Dynamic Range Compression Plug-in A dynamic compressor attenuates the strong sections of a signal in order to reduce the dynamic range of the signal. Most of the compressors used for music or speech production operate on the full-band principle. The entire audio signal is processed via a single gain-control element as described above. A multi-band compressor includes a set of filters that splits the audio signal into two or more frequency bands. Three, four, or five band compressors are perhaps the best compromise between versatility and computational simplicity. After passing through the filters, each frequency band is fed into its own compressing algorithm. After processing, the signals are recombined.
This project employs a tree structured multi-rate filter bank to separate the frequency response of a system into 4 frequency bands so that each band can be compressed separately.
The Project Write-up can be found here - Word Document (2MB) The Project Presentation can be found here - Powerpoint Document (2MB)
Suspended
ELJA - Electric-stringed Jazz Pedagogy Jazz improvisation can be a very complicated or a very simple concept depending on your method to learning and interacting. Some people are auditory learners; some visual or otherwise. This orientation directly affects the method of learning jazz improvisation. This project allows the user to bridge the gap between improvisation conceptualization and the physical stringed instrument. By allowing the instrument to physically show the improvisational concepts, the difficulty in connecting the mental to the physical interface is reduced. This project allows the immediate utility of beginner and intermediate improvisation concepts such as chord/scale relation and common tones; however, future expansion also lends itself to the generation and analysis of more advanced concepts such as motifs and "what you didn't play instead of what you did play."
An electric bass outfitted with LEDs in the neck are controlled by your computer. The user enters into a computer program the tempo, form, and chords of a jazz tune (or selects a predefined one). The program then calculates the appropriate chord/scales for the tune and allows the user to make any desired corrections. When the user agrees on what I call the "improvisation map" the program counts off the tempo and starts the song. One color LED shows the scale for each chord, while another is user-programmable to functions such as highlighting certain notes such as 3rds and 7ths or to show common tones with the next chord. The music that the user plays is sent back to the computer to be recorded for playback and analyzed . Using typical jazz improvisation techniques, the computer analyzes the music and display information to the user such as "right" and "wrong" notes, notes that the user plays too much or little, approximate calculation of the user's 'style' (i.e. bebop blues vs. straightforward 12 bar blues) , or suggestions of artists to listen to based on the user's 'style.' The possible uses of the feedback element to the computer are unbounded.
Please see the following project write-up: ELJA Writeup
Completed
Proposed
MaxiReal - Maximum-gain knob Reality emulator This project will most likely be a VST plug-in to be used by studio engineers to aid them in accurate recording of what the ear actually hears. It takes into consideration the Fletcher Munson curves of hearing perception which I believe is a common cause for inaccurate-sounding recordings of loud instruments such as distorted guitar. Many musicians only like the way their guitar sounds at volume 10. I hypothesize that part of the reason, beyond the physical and electronic overdriving which can in fact be accurately recorded, is because at these high SPL levels our ear reacts differently than if the sound source was soft. A sound recorded at 110+ dB SPL then monitored at 60 dB or less, for instance, will not sound accurate because between the two SPL levels there is no 'ear' mechanism which colors the sound. By applying the Fletcher Munson equalization curves on a DSP, I hope to provide studio engineers with a more accurate recording option.