Depends how you are defining sentiment. Each timeframe would have it's own. Larger timeframe sentiment generally influences moves on the smaller timeframe. Smaller timeframes are where one sees the change in sentiment before observing it on larger timeframes. It's where unusual volume is first detected. To apply sentiment to a single bar - a doji without context would be neutral sentiment. Where the close occurs defines the single bar's sentiment. Where this close occurs in relation to other bars and the volume of those bars would define the series of bars as bullish or bearish. ie. there are 6 - 5m bars in a 30m bar. During that 30m that builds the 30m bar, the 5m bars are either trading within the previous bar's range, testing the limits of that range, Breaking out successfully or failing to follow-through on the BO of that range. The market structure of those 5m bars making HH, HL, LL, LH defines what the 30m bar's sentiment can be prior to the close of the 30m bar.
A measure of sentiment can be seen on the websites of some brokers - they display alongside forex charts the percentages of their clients who are long and short on that particular pair. I don't see supporting information - so we don't know how many clients that means, or what level of total capital is involved, or how often the percentages are calculated etc.. From watching these ratios it seems that intra-day price action doesn't have much effect on the sentiment display. The main factor is the trend on the daily chart. If the pair is in an obvious uptrend, sentiment will be bearish: if its in an obvious downtrend, sentiment will be bullish. You have to remember also that most traders are losing money from their trading accounts, maybe 70-80% - so sentiment is often going to be badly wrong. Some people use sentiment as an indicator to take the opposite direction....