How to add audio to Tensorflow Object Detection API?

[kitkatz]

I'm trying to add audio output for the Tensorflow object detection API on the raspberry pi. For example, if it detects a person from the camera, I'm trying to get the program to say "person". When I use the code below for speech output, I get the error message "TypeError: expected str(string), bytes or os". This is likely because I'm assigning the wrong value to speech. Does anyone know what the right variable/value for the object label?

from subprocess import call
...

speech = classes
call(["espeak", speech)]

- Thanks in advance

[Larz60+]

I get the error message

try (add quotes to make string):
speech = 'classes'
Please always show the entire unaltered error traceback as it contains important process information (use error tags)