In terms of subjectively labeling a language as "stacatto", "lilting", "drawn-out" or whatever, you would have to exclude some languages because they have pre-existing sets of characterizations, for example German is "guttural". There are plenty of unfamiliar languages without a classificatory bias, so that one could expose a population of listeners to recorded examples, and ask them for a word that best describes the sound of the language. Nobody has done this.
The reason why German is said to be "guttural" is that its phonetic inventory has [x] and [ʀ], which are back-of-the-mouth sounds, lacking in (most dialects of) English, though oddly French which also has [ʀ] is not labeled guttural (I suppose it is "nasal"). Certain phonetic segments have distinctive acoustic properties which listeners can easily pick up on, and these things influence the labels that would be assigned.
Norwegian and Swedish are "sing-song", because of the phonetic realization of the pitch accent system. Prosodic properties are probably the most noticeable thing about an unknown language, because the fundamental frequency pattern is the thing that changes most slowly (compared to formant frequencies which change in 10s of milliseconds), and is arguably the "carrier wave" that speech relies on. Apart from changes in fundamental frequency (also amplitude), one would look at the periodic vs. non-periodic vs. silent alternation in the signal, for example how long a span of vowels and sonorants you find, with interruptions for fricative noises, or periods of silence where there is no speech (this itself is detectable as a change in amplitude). This imposes a higher order structure on the speech signal, and you may be able to turn that kind of analysis into Western musical terms like "alegro", "andante" etc.