Manejo de la Mortalidad y Desechos

Neurons orchestrate all muscle movement, managing the delicate co-ordination needed to successfully complete a given task, e.g. walking or lifting something. Similarly, there are neurons controlling the speech-related organs which have to co-operate in the production of speech. Loss of neurons associated with the task of controlling some of the speech-related organs, leads invariably to speech disorders. The speech disorders associated with PD are termed hypokinetic dysarthrias (Darley et al., 1975) and lead to reduced speech intelligibility. They can be broadly categorised as:

1) Hypophonia (reduced voice volume)

20 30 40 50 60 70 80 90 110 120 130 140 150 160 170 180 190 200 210 Age (years) F 0 ( H z ) Fundamental frequency F

0 as a function of age and gender

Males Females

2) Dysphonia (breathy, hoarse voice)

3) Hypokinetic articulation (imprecise articulation due to reduced articulatory movement range)

4) Hypoprosodia/monotonous speech (reduced speech pitch10 variability) 5) Palilalia (dys-fluent or hesitant speech)

The relationship between speech and PD has been studied systematically at least as far back in the 1970‘s by Darley et al. (1969a; 1969b; 1975). They reported reduced loudness, monotonous voice, breathy and hoarse voice, and imprecise articulation in PD subjects. Other studies followed, confirming that PWP exhibit breathiness, hoarseness and articulation problems (Logemann, 1978). More recent studies validate and extend these results comparing healthy controls and PWP (Gamboa et al., 1997), with the majority of patients exhibiting laryngeal tremor during normal or loud phonation (Perez et al., 1996). A 40% reduction in vocal loudness was reported in Fox and Ramig (1997), further endorsed by Ho et al. (2001) who attribute this finding to symptomatic frontostriatal circuit11 dysfunction. Interestingly,

however, the vocal sound pressure level (SPL) during sustained vowel phonation is no different from that of the healthy controls (Rosen et al., 2005). PWP show signs of increased vibrational aperiodicity (the vocal folds‘ oscillating pattern departs from periodicity) and increased breathiness (noise) (Michaelis et al., 1998). A decrease in and variability in speech is also documented (King et al., 1994; Holmes et al., 2000) and these decreases mirror the severity of the disease (Metter and Hanson, 1986). In general, hypophonia and dysphonia

10 _{Pitch is the perceived (i.e. pitch is the psychoacoustic equivalent of the physical measure fundamental}

frequency). It should not be confused with pitch period, which we defined earlier as the inverse of the actual . Pitch can be measured by asking a listener to compare speech signals with a pure sinusoid for which the frequency can be adjusted. The pitch of the original speech signal is then by definition the adjusted frequency of the sinusoid that the listener has determined that gives the same tone output. In general, and pitch correlate well.

11 _{The reader will remember the striatum, part of the basal ganglia (BG). The frontostriatal circuit consists of}

neural pathways connecting the BG with other regions of the brain, which are also involved in higher mental functions.

precede the rest of the disorders (Logemann, 1978; Ho et al., 1998), and 98% of hypokinetic dysarthritic speech pathologies are related to PD (Berry, 1983). PD appears to affect men and women differently with respect to their vocal performance. For example, Cnockaert et al. (2008) report that average increases in male PWP and decreases in female PWP compared to matched healthy controls. Recent studies have not taken into account this male-female distinction, e.g. (Little et al., 2009; Tsanas et al., 2010a), and as we will see in Chapter 7 some measures are very highly correlated with UPDRS in one gender, and almost negligibly correlated in the other, confirming the gender selectivity of PD effects on speech.

Interestingly, a pilot study in 2004 revealed that speech impairment could be detectable as early as five years prior to diagnosis of PD (Harel et al., 2004). The speech recordings of two people (one of whom was eventually diagnosed with PD) of similar demographic characteristics (age, gender, and profession) had been examined for 11 years (including 7 years prior to diagnosis). Although that study consisted in comparing the voices of only two people and the authors caution regarding the interpretation of their findings, it is reasonable to assume that some early symptoms of the disease could be traceable before the patient is diagnosed. Similarly, it is plausible that movement disorders could be discovered prior to diagnosis as well but it is difficult to recruit people into such longitudinal studies. As soon as some dopaminergic cells die, an imperceptibly subtle difference in muscle control might be detectable with sensitive equipment and appropriate tests12. Monitoring speech signals from patients who have been recently diagnosed with PD could facilitate understanding about the progression of the disease and give rise to improved diagnostic and treatment methods.

12 _{As we have noted, typically 60% or more of the dopaminergic cells have already died by the time clinical PD}

Chapter 3

Clinical speech signal processing algorithms

This chapter provides a literature survey of some of the most widely used approaches to processing speech signals in order to extract clinically useful information. Many of these tools originate in the speech signal processing research community and have been developed for various purposes, including speaker identification, speech encoding, and, as is the case with the present study, for extracting clinically-relevant information. Other tools have been developed within distinct but related disciplines such as time series analysis. By definition,

time series refers to a quantity changing in time, and speech falls into this category. Before

surveying the various speech signal processing algorithms, we briefly summarize some recurring mathematical concepts.