Since the invention of Gramophone by Thomas Edison in 1877, vast amountsrnof cultural, entertainment, educational and historical audio recordings havernbeen recorded and stored throughout the world. Through natural aging andrnimproper storage, the recorded signal degrades and loses its information inrnterms of quality and intelligibility. Degradation of audio signals is consideredrnas any unwanted modi cation to the audio signal after it has been recorded.rnThere are di erent degradations a ecting recorded signals on analog storagernmedia. The degradations that are often encountered are clicks, hiss andrn`Wow and Flutter'.rnSeveral researches have been conducted in restoring degraded audio recordings.rnMost of the methods rely on some prior information of the underlyingrndata and the degradation process. The success of these methods heavilyrndepends on the prior information available. When such information is notrnavailable, a model of the underlying undegraded data can be used to generaternsuch prior information. Linear prediction is one of the most widely usedrnmodels to represent speech. However, linear prediction has limitations forrnvoiced speech and music and as such restoration approaches that use linearrnprediction have limited success for voiced speech and music.rnThis research uses recent ndings in linear prediction modeling in thernrestoration of click and `wow and rnutter'. Recent developments in e cientrnalgorithms and computational capability have led to signi cant investigationsrnon the usefulness of `1-norm and `0-norm regularization in the solution tornthe least squares problem. The use of high-order sparse linear prediction forrnovercoming the limitations posed by conventional linear prediction has beenrninvestigated by other researchers. This research investigates the use of highorderrnsparse linear prediction for the detection and restoration of degradedrnarchived audio signals.rnA method is developed that uses the high-order sparse linear predictionrnmodel to estimate the underlying audio signal without priori information onrnthe type of audio and the details of the degradation. The model is then usedrnfor the detection of the degradations as well as for the restoration of therndegraded sample values. The use of the model for two of the most widelyrnencountered degradations in archived audio is investigated.